Professional Documents
Culture Documents
16 Mutamar Kelima
16 Mutamar Kelima
WPS7561
Melissa Adelman
Public Disclosure Authorized
Miguel Székely
Public Disclosure Authorized
Abstract
School dropout is a growing concern in Central America, several underlying factors that affect the marginal ben-
and in Latin America as a whole, because of its conse- efits and costs of staying in school. Global and regional
quences for economic productivity, the inclusiveness of evidence increasingly points to common policy priorities
growth, social cohesion, and increasing youth risks. This for reducing dropout across Central America, including
paper utilizes more than two decades of household survey addressing remaining gaps in access at the pre-primary
data to construct a systematic overview of school dropout and secondary levels, improving the quality of education
at the primary and secondary levels in Central America, at all levels, and strengthening the coverage, targeting,
including recent trends, causes and consequences, and and coherence of existing programs aimed at improv-
evidence on what works to reduce dropout. Within each ing education outcomes. However, additional rigorous
country, poverty, rurality, and indigenous group mem- evaluations, including cost data, are needed to identify
bership are the strongest correlates of dropout, reflecting the most effective specific approaches in each country.
This paper is a product of the Education Global Practice Group. It is part of a larger effort by the World Bank to provide
open access to its research and make a contribution to development policy discussions around the world. Policy Research
Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at madelman@
worldbank.org.
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.
1
The authors thank Pamela Mendoza and José Mola-Avila for their research assistance. The authors thank Pablo
Acosta, Raja Bentaouet-Kattan, Laura Chioda, and Rafael de Hoyos for their valuable comments.
Introduction
Low education quality on average, inequalities in both access and quality within countries, and
risks that affect education for specific subgroups are widely recognized challenges in Central
American education systems, and high dropout rates are a reflection of these challenges. 2 Dropout
is a growing cause of concern in Central America, and in Latin America as a whole, due to its
multiple consequences for economic productivity, the inclusiveness of growth, social cohesion,
and increasing youth risks. In particular, as countries look to accelerate their growth, the
inadequate skills of their workforces present a key constraint (World Bank 2014, 2015a and
2015b). Compared to countries at similar levels of development, countries in Central America are
behind in terms of the educational attainment of their populations. As shown in Figure 1, fewer
young adults in Costa Rica have completed secondary compared to structural peers, other upper-
middle income countries, and the OECD, while Guatemala, Honduras, and Nicaragua all lag
behind the lower-middle income country average.
This paper presents an analysis of school dropout, focusing on secondary dropout, in Central
America, with the goal of providing a systematic overview of the dropout phenomenon, including
recent trends, potential causes and consequences, and what works to reduce dropout. To do this,
we build on substantial recent research in Latin America on dropout and youth idleness (neither
attending school nor working), in particular the conceptual framework presented in Behrman, de
Hoyos, and Szekely (2015), the cross-country cohort analysis by Bentaouet-Kattan and Székely
(2015), and the narrative review by Almeida, Fitzsimons, and Rogers (2015).3
The paper is divided into five sections. Section 1 presents a simple conceptual framework of the
factors involved in dropout as well as the data for the empirical analysis, and discusses the patterns
of dropout over the past several decades, comparisons with the rest of Latin America, and
differences across subgroups within each country. Section 2 documents the economic
consequences of dropping out, while Section 3 turns to the drivers of dropout, presenting empirical
2
Highlighted for instance by Bruns and Luque (2015) and Cabrol and Székely (2013), among others.
3
A related study by Bassi, et.al. (2015) focuses on measuring secondary school dropout and graduation rates for the
countries in the region, by using snapshots of household surveys up to the late 2000s. These authors conclude that
dropout has declined, and point to the increase of school services and in the returns to secondary education as main
causes for the reduction. Other related recent studies that analyze school dropout in the region, although not as the
central focus of their investigation are Cárdenas, et.al. (2015) and Székely and Karver (2015), who look at the
consequences of leaving school prematurely, and particularly the relation between dropout and the percentage of youth
that are neither in school nor working. Another useful reference is the study by UNICEF (2013), which focuses on
school coverage and attendance rates at the preschool and primary levels, for Central America specifically. The study
clearly illustrates that for some of the countries in the sub-region, primary school participation continues to be a
relevant challenge. Previous studies on the education sector of Latin America that include secondary dropout as part
of their analysis (although not as the main focus) are World Bank (2005), and UNESCO (2008).
2
correlations and the available causal evidence. Section 4 considers the global evidence on what
works to reduce dropout, and how this evidence informs the Central American experience. Section
5 concludes.
Global evidence suggests that there are multiple, interacting factors that affect learning,
progression through grades, and ultimate dropout. Diagram 1 presents a simple framework to
organize these factors, built on the lifecycle approach to human capital accumulation. The
framework highlights the foundational theory of human capital, in which individuals decide
whether to persist or drop out of school by weighing the marginal expected costs of continuing
investment in education against the marginal expected benefits of acquiring more years of
schooling (Becker 1967; Behrman, de Hoyos, and Szekely 2015). These costs and benefits are
determined by proximate indicators, which in turn are shaped by underlying individual-
household, community, and macro factors. Early in life, this decision is often taken by parents or
other family members, while later on, decision-making shifts more to the individual, particularly
in families where youth have little parental oversight or support.
In terms of marginal benefits, learning deficits often begin early and grow over time, and
coupled with archaic curricula, contribute to the lack of interest in studying cited by so many
out-of-school youth (discussed in section 3). With relatively weak skill foundations in the early
grades, children increasingly struggle as material becomes more complex in the higher grades,
reducing their perceived benefit to staying in school. In terms of costs, direct costs are generated
by distance to school, lack of seats, and school fees, all of which are issues particularly at the
secondary level across Central America. Families in need of unpaid labor—to care for younger
siblings, do agricultural work, or other activities – face high opportunity costs to school and may
pull children out of school on a regular basis, making it difficult for them to keep up and
progress, eventually leading to permanent dropout. Youth may also see opportunities in the
labor market that are immediately attainable, and decide that earnings in the short term are
preferable to continued investment in schooling. For example, growth in demand for unskilled
labor, in agriculture, factories, or other areas, could pull young people out of school earlier than
they would otherwise leave.
Pregnancy, crime and violence, and other risk indicators can affect both the direct costs and the
opportunity costs of staying in school. Teenage pregnancy rates are particularly high in Central
America, and teen mothers may find it impossible to continue in school, due to lack of needed
supports such as daycare, while other young women may decide that it is preferable to leave
school and start a family early. Similarly, crime and violence are major problems in several
3
Central American countries, with homicide rates for Honduras ranking the highest in the world
and those for Guatemala and El Salvador ranking in the top five. Violence in and around a school
may make it untenable for youth to continue attending, while illicit activity may offer some
youth an attractive alternative to staying in school.
The relative importance of different indicators also shifts over the lifecycle. For example, as
children age into early adolescence, risk factors present in the community are more likely to apply
to them (Cunningham et al 2008). Recruitment into gangs, initiation of risky sexual behavior, and
other risks grow, which can make school itself a more dangerous or unwelcoming place, or provide
temporarily appealing alternatives to remaining in school. At the same time, the prevalence of
unskilled jobs in the labor market may increase the opportunity cost of continuing in school as
youth age.
Many community-level and macro characteristics also help determine costs and benefits. Access
to school, particularly at the upper secondary (US) level, may be very limited in rural communities,
while the quality of schooling services available is often low. Social and cultural norms around
what levels of educational attainment are acceptable, and for whom, also factor into decision-
making, as do the prevalence of teenage pregnancy, early marriage, participation in crime and
violence, and migration, all of which may be both causes and consequences of dropout. Beyond
the community, economic growth and labor market trends both at the national and international
levels affect government budgets and education policies, remittances, and job opportunities, which
in turn affect the quality and accessibility of schooling, its affordability, and its opportunity cost.
These underlying factors are likely to interact within each level, and also across levels in important
ways. For example, economic growth and strong local labor market conditions may increase
household incomes, making more resources available for investing in the human capital of
4
household members. However, the same opportunities might generate the additional effect of
pulling youth out of school by increasing the opportunity cost of remaining enrolled and losing out
on income generation possibilities. These economic changes may also influence the information
set available to youth and their families, shaping their perceptions of the returns to education. The
final result depends in part on the skill content of the increasing labor demand.
At all levels, negative shocks can have substantial impacts on schooling decisions. Health shocks
to individual family members, major weather events that affect entire communities, and aggregate
economic shocks can all impact on the short to medium-term tradeoff of costs and benefits to
staying in school. Importantly, these short-term impacts affect what is often a long-term decision,
as most youth who leave school do not return.
1.2 Data
As the main source of data for our investigation, we rely on household surveys, which have been
systematically held in Central America over the past three decades. 4 As shown in Appendix Table
A.1, the country with the most data is El Salvador, where 19 surveys are available between 1989
and 2013. Panama follows with 17 from 1991 to 2013. For Honduras and Costa Rica, 16 surveys
are available for the periods 1989-2012 and 1987-2013, respectively, while for Guatemala and
Nicaragua, a more limited set of 9 and 7 surveys exist between the 1990s and 2013.
One important advantage of using household survey data is that different education system
structures can be accounted for in order to guarantee comparability in cross-country analysis. A
crucial variable for comparability is the official age at which individuals are expected to attend
each education level, and the number of grades comprising each level. As illustrated in Appendix
Table A.2., most Central American education systems include 12 years of formal schooling
between primary and graduating from upper secondary (US), with entry normally at age 6 and
expected exit at age 17, in line with the rest of Latin America. Two exceptions are Guatemala and
Nicaragua, where both entry and exit are set one year later (as in Brazil). Having access to the
original micro data in household surveys allows adjusting the data accordingly, in order to make
accurate comparisons in attendance by level and official schooling ages.
The data also allow internalizing differences in the definition of types of secondary schooling and
relating dropout to important individual and household characteristics. In all six countries in
Central America, the post-primary and pre-higher education cycle—which we generally label as
secondary education (secondary) here—is split into two segments. The first segment of lower
secondary (LS) has a typical duration of 3 years, followed by another 3-year segment of US. Apart
from allowing for consistent comparisons, relating ages with schooling levels is particularly
4
An alternative followed by other studies has been to rely on official administrative records from the schooling system
itself. However, these types of data bases do not usually include information of those not attending education services,
which are a key component for our analysis.
5
important for normative reasons. While primary education is already compulsory in Central
America, there are differences thereafter. The main one is that LS is also compulsory in Costa
Rica, Guatemala, El Salvador, and Panama, as well as Honduras as of 2013, but not in Nicaragua
(OEI 2010).5 Another important advantage is that since household surveys incorporate information
on a wide set of personal and household characteristics, it is possible to differentiate school dropout
patterns for different population subgroups or family situations. Furthermore, the data incorporate
other variables such as labor market conditions that are useful for understanding schooling
decisions.
Yet another advantage, which will be illustrated in the following sections, is that household
surveys include representative samples of all age groups because they are normally structured
around the demographics of each country. This allows us to follow the trajectory of different
cohorts over time. The case of El Salvador, where 19 observations are available over a 24-year
period, illustrates this. If one identifies all individuals ages 1-5 in the first survey of 1989, it is
possible to locate the same group six years later in the 1995 survey, when they are ages 7-11,
precisely the ages at which children are expected to be enrolled in primary in the country. It is
therefore possible to identify the proportion actually in school to compute the net coverage of
education services at this level. Six years later, in the year 2001, the same cohort can be identified
again at ages 13-17, which is the age group expected to be attending secondary. Within the cohort,
it is also possible to verify whether those ages 13-15 are actually enrolled in LS as expected, and
if the older half of the cohort who are ages 16-17 are enrolled in US. With this information, net
enrollment in both segments can be determined, as well as the share that is enrolled at a lower level
(and therefore overage) and the share that is out of school.
The following sub-sections implement this approach in order to identify patterns of secondary, LS,
and US dropout over time in Central America, with comparisons to the rest of the region.
1.3 The general context: Progress in education in Central America over six decades
Taking a long-term regional view of the extent and speed of education progress helps to place the
current dropout challenge in perspective. This is illustrated in Figure 2, which plots the average
years of schooling of individuals belonging to different cohorts across all of Latin America,
reported in the latest household survey available for each country (circa 2013). According to these
estimations, individuals belonging to the cohort born in 1935-40, and who were ages 78-82 in the
5
Some countries in Latin America with slightly different arrangements are Bolivia, Chile and the Dominican Republic
where LS has a 2-year duration, while in Colombia and Peru secondary comprises 5 years and is not divided into two
segments.
6
year of observation (that is, 2013), attained on average 4.7 years of schooling. 6 When comparing
them to the average of 11.5 years of schooling attained by the cohort born 60 years later in 1990-
1995, and who were ages 18-23 in 2013, we conclude that that there has been an increase of 6.8
years of average educational attainment across generations over the period.7
This way of plotting the data is useful for viewing long-term trends, and also for verifying the
moment when different benchmarks were reached. For example, as the figure shows, the cohort
born around 1950—who turned 18 in 1968 and 63 in 2013—attained 6.5 years of schooling, which
is equivalent to having completed the primary level. Our interpretation (subject to the caveats
mentioned in footnote 7) is that already in 1968, the new entrants into the labor force in Latin
America had reached a first phase of evolution where the average individual had attained basic
education. According to the data, a second benchmark was reached in 1993, when the cohort
turning 18 completed on average 9.2 years of school, which is equivalent to having completed LS
in the education systems of most countries in the region. The value of the last point to the right in
the figure shows that the cohort turning 18 in 2013 completed 11.5 years of schooling on average,
short of the 12 years that would correspond to completing US. This indicates that across Latin
America, the “average” individual age 18-22 has actually obtained some US education, but
dropped out before graduating.
The overall trend in Latin America helps to place Central America in context, and to identify the
diversity within the sub-region. From this long-term perspective, there are clear differences
between Costa Rica and Panama on the one hand, and Guatemala, Honduras, and Nicaragua on
the other, with El Salvador somewhere in between. Figure 3a shows that Panama and Costa Rica
reached the first benchmark of average primary completion around 15 years earlier than the Latin
America average. Costa Rica levelled off with Latin America by reaching the second stage also in
6
The year of birth of each cohort appears at the bottom of the Figure. For illustration, take again the case of El
Salvador, where the latest household survey was held in 2013. In these data it is possible to identify, for instance, all
individuals in the 78-82 age group, who by definition were born between 1935 and 1940, and thus turned 18 during
the period 1953-58 -which is the age at which individuals are expected to have exited US and either continued to
tertiary or entered the workforce. Since household surveys report the number of official years of schooling and the
maximum level reached for all individuals, it is possible to determine the average schooling level of the cohort.
Additionally, since schooling occurs at early ages and typically stops by ages 20-24, it is possible to interpret the result
as the education acquired by the cohort during schooling age. The same exercise can be performed for all individuals
who were age 18-22 in 2013, and were therefore born in the 1990-1995 period, as well as for all cohorts in between
the two extremes. The Figure plots the average years of schooling for subsequent cohorts born between 1935 and 1995
for the 18 Latin American countries for which we have data –the data points correspond to unweighted regional
averages.
7
Evidently, cohorts can change in composition over time and this can affect the averages obtained. Specifically, if
there is differential mortality, emigration, or immigration, the averages will not reflect the profile of the same
population over time, and can bias the picture. For instance, if those migrating outside of the country at prime age
have relatively lower or higher education levels that those remaining in the country, the average years of schooling
of those observed in the remaining cohort might be biased upward or downward, respectively. Unfortunately there is
not sufficient data to correct for this for the countries included in our analysis, but we believe that the picture is still
illustrative. In the case of mortality, if rates are higher the lower the education level, pictures such as that in Figure 2
would be upward biased.
7
1993, but has recently lost some ground—18 year olds in 2013 had attained an average of only
10.9 years of education. Panama reached the second stage of LS completion around 20 years before
the Latin America average, but subsequently levelled off to the point that the latest cohort registers
the same average years of schooling as the average for the region. El Salvador shows the opposite
pattern, lagging behind the Latin America average by about 10 years in reaching the first two
benchmarks, and then increasing faster and almost catching up, to the extent that the latest cohort
reaching US exit age falls short only about 0.6 years of schooling with respect to the region –with
an average of 10.9 years completed, the same as Costa Rica. Thus, for Costa Rica, Panama, and
El Salvador, the US dropout phenomenon is clearly a central challenge.
Guatemala, Honduras, and Nicaragua also face this challenge, but additionally, they are still
struggling with the issue of LS completion. As shown in Figure 3b, Guatemala and Honduras
reached average primary school completion about 30 and 25 years after the Latin America average,
respectively, and have just entered the second phase where the average individual of US exit age
completes LS. Nicaragua reached the first benchmark more quickly (10 years later than the Latin
America average), but also only recently reached the second benchmark of average LS completion.
For every country, national averages mask large within-country disparities, which will be explored
in the subsequent sections. The picture that emerges is one of a sub-region facing new challenges
of the 21st century, while still dealing with unresolved issues from the past.
By following a similar cohort approach, Bentaouet-Kattan and Székely (2015) have shown that
higher enrollment and completion rates at the primary and LS levels in Latin America have implied
higher dropout rates at the US level. In Figures 4a and 4b, we use the same approach as those
authors by following the enrollment trajectories of two cohorts, the first born 1980-1982 and the
second born 1994-1996.8
Across Central American countries, a variety of patterns is observed. As shown in Figure 4a, the
dropout rate in Costa Rica during LS ages fell by 11 percentage points across cohorts, dropout
during the transition ages between LS and US remained practically the same, and dropout during
US ages increased by 4 points.9 Therefore, for secondary education as a whole, there was a decline
in dropout by 6 percentage points. Panama in the same figure follows a less similar pattern, with a
slight reduction in dropout during LS ages, small changes during transition, and a small increase
8
See Appendix Table A.3 for a complementary measure of dropout, specifically looking at the educational
attainment of 25-29 year olds in each country’s most recent household survey.
9
In this section, enrollment and dropout are considered from the perspective of what schooling level a child should
be attending given their age – the issue of overage is examined in Section 1.6 below.
8
during US ages. The result is a practically unchanged overall secondary dropout rate between
cohorts.
The trajectories of the same two cohorts for El Salvador are shown in Figure 4b. Here, dropout
during LS ages and during transition remained practically the same, but there is a considerable
increase in dropout at US ages of 10 percentage points, which results in an increase from to 30 to
39 points in the overall dropout rate of youth of secondary school age. In Guatemala and Honduras,
the overall secondary dropout rate declined, as greater declines at LS ages and transition
compensated for increases at US ages. In Nicaragua, overall secondary dropout remained
practically constant, although with a pattern opposite to that observed in the other five countries.
In Nicaragua, LS and transition dropout both increased slightly, but the effect was counterbalanced
by a reduction in dropout at US ages of 3 percentage points.
Table 1 summarizes the results for the latest cohort for Central America, and allows characterizing
the different country patterns currently observed. In Costa Rica, Panama, and El Salvador, the
majority of secondary dropout occurs during US ages. In Guatemala, Honduras, and Nicaragua
somewhat similar proportions drop out during LS and US ages. In these last three countries, the
emerging US dropout challenge is combined with still significant shares of youth leaving school
prematurely during LS. In these same countries, it is important to note that about 20 percent of
the younger cohort (those born 1994-1996) was out of school by the age of transition to LS,
meaning that primary dropout, and in some cases even initial participation, also remain a challenge
in these countries.
A similar analysis can be performed separately for each gender. Figure 5 summarizes the change
in attendance at each stage for males and females separately, and shows that dropout is slightly
lower for females than for males during LS, with the exceptions of Guatemala and El Salvador. In
the transition between LS and US, female dropout is also lower in Costa Rica and Panama, while
in the other four countries practically the same rate is observed for both genders. For dropout
during US, a somewhat different picture emerges. In Costa Rica, Guatemala, and Honduras
dropout rates across gender are basically equivalent, while in Panama and El Salvador female
dropout rates are higher. Nicaragua is the only country where male US dropout is considerably
higher, with a 5 percentage point difference between the two.
Cohort analysis such as that presented above helps to clarify the dynamic aspects of dropout that
underlie the long-term changes shown in Figure 3. In order to complete the picture in terms of
differences across and within countries, we also describe current levels of enrollment at different
ages. Figure 6 plots the proportion of individuals ages 7-18 who were enrolled in the schooling
system circa 2013. The main feature of the figure is that between ages 7 and 11, relatively small
9
differences of around 10 percentage points are observed across the six Central American countries,
with Nicaragua at the lower end of the spectrum, and Costa Rica and Panama at the upper end. 10
However, at age 12, when secondary entry is generally expected to occur, considerable and
expanding disparities arise, to the extent that by age 17 the difference between the highest
enrollment rates observed in Costa Rica and Panama (75 percent) are practically 30 percentage
points higher than that observed in Honduras (46 percent). In between is El Salvador with an
enrollment rate of 65 percent of 17 year olds, followed by Nicaragua with 60 percent, and
Guatemala with 51 percent.
As already mentioned, one advantage of household surveys for the present analysis is that they
allow identifying within-country differences. Figure 7 plots the average differences in enrollment
rates for 12-17 year olds between females and males, indigenous and non-indigenous groups, rural
and urban areas, and individuals belonging to households in the poorest and richest quintiles of the
household income distribution.11 These underlying factors, as presented in the conceptual
framework, affect the marginal costs and benefits of staying in school, and therefore are expected
to be correlated with dropout.
The largest differences in enrollment rates in Costa Rica, Panama, El Salvador, and Guatemala are
observed between youth belonging to households in the richest and poorest quintiles. In Honduras
and Nicaragua, more pronounced differences are observed when the population is classified by its
urban/rural location, which in addition to relating to income, is likely related to the supply of
secondary school services. Disparities across indigenous and non-indigenous groups are sizable in
Panama, Guatemala, and Honduras. An important result is that gender differences are relatively
small, with females generally showing slightly higher enrollment rates than males across the
secondary age group, with the exceptions of El Salvador and Guatemala where males show slightly
higher rates.
In most Central American countries, schooling enrollment disparities across the household income
distribution are relatively small in primary and start to expand in LS. As Figure 8 shows, at the
primary level (first panel in the left) more than 90 percent of children in the relevant age group are
actually enrolled in primary across the income distribution in most countries. The exceptions are
Guatemala and Nicaragua, where enrollment rates for the poorest four deciles are between 80 and
90 percent. Enrollment rates reach over 95 percent in all countries as household income levels
10
Some recent studies for Nicaragua (see for instance Ministerio de Educación de Nicaragua (2014)) and Guatemala
(Ministerio de Educación de Guatemala (2014) have argued that education coverage rates at the primary level after
2010 have declined, although there is controversy about the data sources from which this conclusion emerges. In
particular, the calculations are based on school attendance rates from the administrative records of the Education
Ministry while data on the number of individuals in the relevant age group (the denominator in the equation) is from
extrapolations based on Census data. We do not find the same pattern in our household survey data.
11
Country-specific plots are in Appendix Figures A.1 to A.4.
10
rise. In contrast, LS enrollment rates among individuals in the poorest decile of households are
about 30 percentage points lower than those in the richest decile, with extreme cases like Honduras
registering a distance of more than 50 points. By US, differentials between the poorest and richest
decile reach 40 percentage points in most countries, with Panama showing the largest gap of 55
points. These results are consistent with the framework, as poorer children may estimate lower
benefits from continued enrollment, and face relatively higher direct and opportunity costs
particularly at the secondary level.
Access to public education services reflects these same differences.12 As shown in Figure 9, of all
individuals enrolled in primary public schools, a majority live in households at the bottom of the
income distribution. Between 12 and 17 percent of public primary students are from households
in the first decile, while only 6 percent or fewer are from the richest 10 percent of households.
These results would suggest that, in general, primary services are progressive, benefitting the poor
more than the rest. Along the same lines, public LS services could be characterized as focusing
on the poor and middle classes, since the shares of students living in households in deciles 1 to 7
are somewhat similar, while fewer than 6 percent of those enrolled belong to households in the
richest decile. In contrast, US services appear to be regressive, since the proportion benefitting
from these services generally increases along the income distribution up to decile 9. Extreme cases
are Nicaragua and Honduras, where the proportion of students enrolled in public US services is
three times larger in decile 9 compared to decile 1. In Costa Rica and El Salvador, where the
differentials are less pronounced, the proportion of all US public students that belong to the poorest
and richest households, respectively, is practically the same. The role of access in dropout will be
discussed further in Section 3 below.
The previous discussion described school participation based on which levels individuals would
be expected to be enrolled in at each age. Figure 10 summarizes the data on “net coverage” rates
from the most recent household surveys, which refer to the proportion of individuals actually
enrolled at the level that corresponds to their age, according to the structure of each country’s
education system. In general, these rates are quite high in primary, but decline at LS and especially
at US. For US, the highest rate is observed in Panama, where 53 percent of youth in the 15-17 age
group is actually enrolled at this level. In Honduras and Guatemala, only 27 and 20 percent of this
age group, respectively, are enrolled in US.
These results reveal that there are considerable numbers of US age youth in Central America who
are still enrolled in school but overage, and not enrolled at the US level as would be expected. This
is illustrated in Figure 11, were each country is compared to the average of Latin America, where
12
Data for Panama is not presented here, since the household survey does not allow identifying whether attendance is
in public or private schools.
11
at US age, 24 percent are already out of the schooling system, 6 percent are still attending primary,
37 percent are attending LS, and only 33 percent are actually attending US as expected. The only
country in Central America with better indicators is Costa Rica, where overall enrollment is higher:
only 17 percent are out of school, and the proportion of US age individuals actually attending the
expected level is 40 percent—7 points above the Latin America average. Panama shows a similar
rate of overage as the Latin America average, but has lower proportions out of school and enrolled
in US. El Salvador shows a different picture. In this case, overage is lower (at 26 percent) and
enrollment in US is higher (37 percent), but the proportion out of school (38 percent) is
significantly above the Latin America average. Honduras, Guatemala, and Nicaragua show a
similar pattern of net coverage rates below the regional average, combined with significantly
higher shares out of school that reach between 43 and 47 percent, twice as high as the Latin
America average. Thus, in these countries, overage is lower, but only because more youth have
already left the schooling system by age 17.
Students who are overage due to starting late and/or repeating grades are more likely to drop out,
as demonstrated with administrative data by Manacorda (2006) in Uruguay and Branson et al
(2014) in South Africa. This relationship may exist because the same factors that drive late school
entry or poor performance also drive later dropout, because the opportunity costs of remaining in
school rise as children age, because overage and poorly performing students are pushed out of the
system rather than supported, or any combination of these and other factors. Regardless of the
underlying drivers, the existence of a correlation suggests that the results in Figure 11 are an
indication of the prevalence of dropout in Central America in the near future.
The previous section documented the patterns of school enrollment and dropout at the critical ages
where individuals would be expected to be in secondary school. In this section, we explore some
of the consequences of prematurely leaving school.
Leaving the schooling system before US completion can have a wide array of effects. Here we
explore some of its labor market consequences, which depend on the interaction between the
accumulation of human capital acquired through education, and the characteristics and evolution
of the productive sectors of the economy that generate the conditions for its use. For instance, if
there are restricted employment opportunities, the actual effect of early exit on economic well-
being might be lower than expected.13
13
In Central America, and across Latin America more broadly, there are some signs that although education can make
an important difference, its potential is not being fully capitalized precisely because of the slow generation of
opportunities to put it to work. One clear illustration of this is that the proportion of youth in Latin America who are
out of the schooling system and are not employed, is high and has been persistent over the years -this has been
illustrated by Cardenas et.al. (2015) and Székely and Karver (2015).
12
The probabilities of working, and of working in the formal sector, are strongly related to
educational attainment throughout individuals’ working lives. As shown in Table 2,
unemployment rates among 25-45 year olds are higher among individuals who did not complete
US, as compared to the population average and to those who attained 13 years of schooling or
more. The largest differences are found in Costa Rica, where individuals with 13 years of schooling
have unemployment levels 5.4 percentage points lower than those who attained fewer years.
Informality is also more prevalent among workers with lower education levels. In Panama,
Guatemala, Honduras, and Nicaragua the rate of informality among workers who only reached LS
is over 32 percentage points higher than for workers with at least some higher education. The
extreme case is Honduras, where the difference is 46 percentage points.
There are also important differences in wage levels, especially for individuals who are able to
graduate from US and enter higher education. Table 3 presents estimates of relative returns to
completing a full course of secondary education, and to entering higher education, with respect to
primary. In the case of the relative returns to secondary completion, the most recent data show that
with the exception of Guatemala, all Central American countries register higher levels than the
Latin America average. However, only Panama and Guatemala –and Costa Rica to a more limited
extent—share the regional declining trend in this variable since circa 2000.14 This suggests that in
most of Central America, secondary completion continues to generate substantial income
dividends on average, making it costly to leave before completing US.
The relative returns to higher education in Central America show that completing US and
continuing to the next level also has substantial economic rewards. In El Salvador, and to a more
limited extent Panama, such rewards are substantially larger than what is observed in Latin
America. Even though Costa Rica, Guatemala, and Honduras register a declining trend during the
2000s that is more pronounced than the slight reduction in the regional average, returns to this
level continue to be significant. In El Salvador and Panama, relative returns to higher education
have continued to rise substantially during the past decade.
An interesting result for Central America is that the economic gains of completing US are
considerable compared to completing only LS. Figure 12 illustrates this by plotting the difference
in average wages at ages 19-24 that result from completing each segment of secondary. For
example, when comparing wages for those who only completed primary versus those who
completed LS, differences range from only 2, 5, and 6 percent in Nicaragua, El Salvador, and
Honduras, to 15 and 25 percent in Guatemala and Costa Rica. However, when comparing the
average wages of those who completed US versus those who reached a maximum of LS,
14
Returns to secondary and tertiary completion in Costa Rica rose substantially between 2007 and 2013, potentially
reflecting post-crisis improvements in the economy (World Bank 2015a).
13
differentials go from 22-24 percent in El Salvador and Guatemala, to between 39 and 46 percent
in Panama, Nicaragua, and Honduras, and reach 75 percent in Costa Rica.
While average returns to completing US and beyond remain substantial, within each country there
may be a wide dispersion of returns that depends on many factors. Individuals who receive a
higher quality and more labor-market relevant education, who live in areas with stronger labor
markets, who have stronger social networks, and so on, are likely to reap larger rewards from their
human capital investment. For example, in Costa Rica, tertiary training graduates receive on
average anywhere from negative 6 percent to positive 9 percent returns depending on the
institution attended and controlling for individual characteristics (World Bank 2015a). Similarly,
Gonzalez-Velosa et al (2015) find that substantial shares of tertiary graduates in Chile and
Colombia receive a negative return on their investment, reaching a disturbing 59 percent of
graduates from technical and technological institutions in Colombia. This dispersion of returns
likely informs young people’s decisions about whether to persist in school, and reflects (at least in
part) variation in quality within each country’s education system, an issue discussed further below.
Beyond individual labor market consequences, dropout can have other important consequences.
A growing body of evidence primarily from the United States indicates that higher levels of
educational attainment causally impacts health, parenting, and other aspects of life, producing both
private and social benefits (Oreopoulos and Salvanes 2011). Across several countries where
studies have been completed, perpetrators of crime, mostly men, have lower levels of completed
education than the general population, and in a cross-country panel analysis, the rate of secondary
completion in particular is strongly negatively correlated to violent crime rates (Chioda 2013).
Chioda (2013) reviews the evidence from well-identified studies of the effects of increased
schooling on crime, showing that measures that increase time spent in school, years of schooling,
and the quality of schooling can all reduce crime in the short and long run.
As shown in Table 1, countries in Central America differ in the segment of secondary where
dropout takes place. Therefore, as discussed in the conceptual framework above, the reasons for
dropout are likely to vary from country to country, as well as along the schooling trajectory. A
relevant source of information for understanding the causes of dropout is found in the same
household surveys used so far, as most of them directly ask youth who have left the education
system to register the main reasons why they decided to leave school.15
15
It should be mentioned that these are self-reported causes, and that a multiple choice format with pre-defined
alternatives is provided for the answer. All possible reasons for dropout are not included, and the response is based on
each young person´s perception of the cause of dropout, and also on their understanding of the meaning of the multiple
choices provided for the answer.
14
Figure 13 summarizes the self-reported causes for individuals of secondary school attendance age
for the five Central American countries for which household surveys report this indicator. We have
grouped the answers into five categories, comprising economic factors (related to having to work,
insufficient family or personal income for investing in education, not enough resources for
travelling to school or to purchase materials, etc., and poverty), personal factors (sickness, having
to perform household responsibilities, pregnancy, marriage, etc.), not interested in studying (which
includes categories such as “dislikes school”, “considers what schools teach to be irrelevant”,
“thinks that what is learned in school is not relevant”, etc.), no access to education services, and
other reasons that are not included in the previous categories.
In four of the five countries, the reason that most youth indicate as the primary cause for having
dropped out is that they were not interested in studying: 37, 40, 41, and 47 percent report this as
the main cause in Costa Rica, El Salvador, Nicaragua, and Panama, respectively. The second
leading cause cited by youth is the lack of economic resources to continue attending school, with
percentages around 25 percent in all five countries. Honduras stands out as the only country where
the majority of youth reports economic causes as the main factor, with 53 percent, while only 29
percent express that lack of interest in studying is the main reason.
The importance of personal reasons varies from 11 percent in Costa Rica, to 20-21 percent in
Panama and El Salvador. Access to education services is reported as the main cause by a relatively
low share: in Nicaragua, 8 percent of youth out of school report that the reason why they do not
attend is because the service is not available, while in Costa Rica, Panama, El Salvador, and
Honduras the proportions are 6, 5, 4, and 2 percent, respectively. However, it is possible that this
reason is correlated with economic factors, since access to education services can be associated
with the economic capacity of individuals and households for financing transport costs to where
the service is provided.
The declared reasons for dropping out also differ by gender. For males, economic reasons are most
often cited, while females are more likely to cite personal reasons, including having to perform
household tasks (see Figure 14). In most countries, a larger share of males declare lack of interest
as the main cause for having left school, although this category is highly relevant for females as
well. In order to go beyond these individual perceptions, the following sections examine empirical
correlations and causal evidence on dropouts, utilizing the structure of the conceptual framework
presented above.
15
by weather shocks in particular, have been exploited in Central America, the broader Latin
America region, and globally – insights gained from these will be mentioned below. In Section 4,
evidence from randomized evaluations of interventions aimed particularly at reducing dropout will
be discussed.
Given the limited causal evidence, we focus in this section on verifying the existence and
magnitude of correlations between dropout and the factors identified by the conceptual framework.
In particular, we estimate the marginal effect of variables captured in household surveys on the
probability of being enrolled in school in each country for youth ages 15 to 17.16 We estimate a
probit regression for each country using the latest household survey available, first focusing on
individual (X), household (H), and community (Z) characteristics:
𝑒𝑛𝑟𝑜𝑙𝑙𝑒𝑑𝑖 = 𝛼 + 𝛽1 𝑋𝑖 + 𝛽2 𝐻𝑖 + 𝛽3 𝑍𝑖 + 𝜀𝑖 (1)
Figure 15 plots the values of the marginal effects of gender and belonging to an indigenous group
on the probability of attending, while controlling for all variables included in the specification, and
confirms that in general, being female is positively correlated with being enrolled at secondary
ages.17 In El Salvador and Guatemala, however, the opposite is the case. This result is consistent
with evaluation results from conditional cash transfer programs that have found that the gender
gap in school enrollment has virtually disappeared in Latin America (Fiszbein, et.al. 2009).
Belonging to an indigenous group is in all cases negatively correlated with enrollment, with
significant and sizeable effects in Guatemala, Honduras, Nicaragua, and Panama, with marginal
negative values of 6 percent and below. This result can be signaling several different underlying
relations, but since it results from controlling for household income and rural location, it may
reflect other elements not included in the equation, such as the relevance and quality of education,
including access to bilingual education or the adaptation of national curriculums to local contexts,
which are important issues in Panama and other countries (World Bank 2014a).18
The marginal effects of the second set of variables related to household characteristics is presented
in Figure 16. In all countries, household poverty, an unemployed household head, and the youth
playing the role of main breadwinner, are each negatively correlated with enrollment, but statistical
16
This is an approximation for dropout, as a small share of the youth who are not enrolled were never enrolled, and
another small share may return to the schooling system in the future.
17
The (*) in this and the following figures denotes that the marginal effect is statistically significant at the 90 per
cent level or above. Table A.3 presents the regression coefficients and standard errors depicted in Figures 15 through
17.
18
Indigenous group status is also self-reported, which may affect how these correlations are interpreted.
16
significance varies. The negative effect of poverty is significant across countries, and is larger in
Guatemala, Honduras, and Nicaragua.
We include in the estimation the share of household income that originates through remittances,
in order to capture one of the channels through which migration, a particularly relevant factor in
Central America, can influence enrollment. Among the five countries (all but Panama) for which
data are available, remittances are significantly positively correlated with the probability of youth
being enrolled in school at ages 12 to 17 in all but Costa Rica. The effects are particularly large in
Honduras and El Salvador, with positive values of 11 and 9 percent, respectively.
The marginal effects of additional characteristics of the household are plotted in Figure 17. In
particular, we include a dummy variable for single-parent households which may relate to the
amount of family support youth receive and what they are responsible for in the household. This
may also be a mechanism through which migration affects dropout, since when an important
breadwinner leaves the family unit, responsibilities are likely reassigned, with new demands on
income-earning activities or household-related chores that may impose constraints on time use by
secondary age youth, which can affect school attendance.19 The marginal effect on school
enrollment of being in a single-parent household is negative in every country and significant in
most cases.
If single-parenthood is associated with migration, the value of the marginal effect for this variable
should be added to the positive effect of remittances to obtain a more complete sense of the overall
effect. When doing so, we find that the positive effect of remittances on enrollment is canceled out
in Costa Rica and Guatemala (although marginally). In the other countries where both effects can
be measured, the positive effect prevails. These positive correlations, however, have not held up
in research for El Salvador that attempts to control for a more complete set of factors that may
affect both remittances and school participation through various techniques (Propensity Score
Matching, Instrumental Variables, etc.) and by using panel data (Acosta 2011a; Acosta 2011b). In
these papers, the author finds that both the gender of the migrant and the gender of the remaining
youth matters – for example, while male migration does not significantly affect enrollment, female
migration reduces it even though women send more remittances, potentially in line with the
discussion of reassignment of household duties discussed above.
In general, we find that higher education levels of the household head tend to have a positive
marginal effect on enrollment, but, holding other factors constant, is not significant in most
19
Evidently, there are other reasons why a household can be of single parent, but due to the high migration rates in
some CA counties it is likely that an important source is precisely the exit of one parent that seeks employment in
another region or country.
17
countries.20 This variable is of interest for several reasons. One is that parental education can
influence how much parents invest in their children’s human capital development from the
beginning of life onwards. Additionally, as in the case of household income, it can be an indicator
of the level of economic mobility in society, where a lower influence of parental education on
children´s schooling would be an indication of greater mobility. To explore this further, we
compute the unconditional correlation between parent´s education and the value of the schooling
gap, measured as the difference between the number of years of schooling expected given the age
of an individual and the actual number of years completed, for all individuals in the 12-17 age
range (Ferreira et al 2012).
Table 4 presents the results for each Central American country for circa 2000, 2005, and 2012. For
each of these periods, the Latin American average is below the result for Central America, which
suggests that in general, (schooling) mobility is lower in Central America. The countries with the
highest correlation coefficients are Honduras and Nicaragua with values of 51 percent, suggesting
that these two countries have the lowest educational mobility in the sub-region. Guatemala,
Panama, and El Salvador follow with values of 47, 46, and 44 percent, respectively, and Costa
Rica registers the lowest value with 40 percent. As for the evolution of these relationships,
Guatemala is the country with the largest decline over time, followed by Panama, El Salvador, and
Costa Rica. In contrast, Honduras and Nicaragua, show constant correlations since the year 2000,
suggesting no progress in mobility.
The last set of marginal effects in Figure 17 focuses on the location of households in rural versus
urban areas. The estimates show that in all countries, living in a rural area is significantly correlated
with the probability of being enrolled during secondary age, and notably, this is the marginal effect
with the greatest magnitude in El Salvador, Guatemala, Honduras, and Nicaragua, where living in
a rural area is associated with a reduced probability of enrolment of 13, 15, 16, and 14 percent,
respectively. There are several channels through which geographic location can influence school
attendance, and given that this effect results from an estimation that controls for personal and
household characteristics, it is possible that this is capturing the effect of access to services,
especially at the US levels. For example, in Honduras, analysis finds that over 1,000 additional
secondary schools in rural areas would be needed to provide access for all secondary-age youth
who are currently out of school, compared to the fewer than 200 that are currently operating today
(World Bank 2014b).
The correlation of service supply with enrollment can be directly examined in the case of El
Salvador, where we have access to detailed information at the municipality level on the number of
schools of each level from the 2011 National School Census, which we merge with the household
survey. Table 5 presents our results and compares them with the base estimation from which the
20
The lack of significance is not entirely surprising given how many other variables related to household head’s
education are being controlled for, including household poverty and single-parent status
18
coefficients in Figures 15-17 are derived. Two additional estimations are presented. The first
includes only the number of schools for each level in the municipality of reference, and its effect
is not statistically significant. However, in the second estimation (third column of the table) we
compute the number of schools per number of individuals of the relevant school age in each
municipality, and the result is that the marginal effect is statistically significant and quite relevant
in magnitude.21 According to our estimates, there is a negative marginal effect of almost 14 percent
from having fewer schools per school age individuals, and it is notable that the magnitude of the
coefficient for rural location falls significantly –to about one half of its value- when incorporating
this variable. In this estimation, the marginal effect of the supply of education services actually
shows the largest influence on school attendance as compared to the rest of the variables
considered. This suggests that the availability of services is still an important determinant of
enrollment, at least in countries like El Salvador and Honduras.
As discussed in the conceptual framework, shocks at the household, community, or macro level
can also play an important and lasting role in schooling decisions. In an assessment of global
evidence, Baez and Santos (2010) find that natural disasters impact educational attainment through
multiple channels, and have negative and durable net effects. Disasters can reduce household
income and assets, destroy productive infrastructure, and change local labor markets, all of which
directly impact on educational choices. In Central America, the 2001 earthquakes in El Salvador,
Hurricane Mitch in Nicaragua, and Tropical Storm Agatha have been associated with higher rates
and intensity of child labor following the disasters, and, in the case of Agatha, removal of children
from school to facilitate this work and/or cut back on schooling-related expenditures (Santos 2007;
Baez et al 2007; Baez and Santos 2008; Baez et al 2015). Depending on the context, effects may
be heterogeneous across children within a family: Bustelo (2015) finds that young adolescents
(ages 13-15) increased their workload while younger children did not, and that only boys’ school
participation was affected after Tropical Storm Stan hit Guatemala. Importantly, Baez and Santos
(2010) find that the largest impacts are often on the poor.
21
The population data is taken from the National Population Census of 2007 and extrapolated to 2012.
22
For a larger list of references on analysis of the causes of school dropout see Bentaouet-Kattan and Székely (2015).
Other related analysis focusing on specific aspects of early dropout are Cardoso and Vermer (2007), South, et-al.
(2005), Ingrum, (2006), Marcotte (2013), Dearden, et.al. (2008), Mo, et.al. (2014), Bentaouet-Kattan and Székely
(2015b), Chioda (2013), Ibañez, et.al. (2013), and Munyo (2013). Useful references that are specific to CA countries
are Jimenez and Gaete (2010) for Costa Rica, Marshall, et.al. (2014) for Honduras, and PREAL (2014) for Nicaragua.
19
are taken. The authors present a base regression estimated through differences in differences which
accounts for potential problems of omitted variables bias and endogeneity, among other issues,
that can be used here for further exploration. The general specification used is:
Where changes in enrollment for cohort i between periods t and t-1 for country c are a function of
a set of independent variables observed for each cohort for each period of time, for each country.
Four groups of variables are considered. The first group (D) includes the population fertility rate
in the year of birth of each cohort and the share of LS students belonging to the poorest 40 percent
of the population, which are classified as “demand” variables, since they intend to capture to some
extent the pressure on education services stemming from population growth and from the increase
in the number of entrants from poorer backgrounds into the schooling system. The second group
(E) includes the rate of growth of per capita Gross Domestic Product (GDP) and the rate of
inflation in the economy, as indicators of general economic and stability conditions. The third
group (L) incorporates estimates of the returns to secondary and higher education, relative to
primary, as measures of the expected marginal benefits for continuing in the education system
until at least US completion. Finally, the fourth group (R) captures labor market opportunities
through the average employment rates and wage levels of individuals age 25-45.
Bentaouet-Kattan and Székely’s results support the conclusion that the lack of progress in reducing
school dropout at the US level in Latin America is the product of countervailing forces. Increases
in the share of poorer youth accessing the schooling system combined with reductions in the
relative returns to education, especially at the US level, have had a dropout-increasing effect.
However, this effect has been counteracted to a large extent by the positive effects on enrollment
from lower fertility rates, greater economic stability, higher wages, and more employment
opportunities, which could be promoting increased human capital investments through an income
effect.
In order to explore the predictions of the model across Central America, we use the regression
coefficients in the base regression of Table 3 in Bentaouet-Kattan and Székely (2015), and apply
the changes in the independent variables of each of the six countries to verify the influence of each
on the change in the rate of enrollment within each cohort over time, by using all the data available
for each country, estimated at the mean. Table 6 presents the trend in each of the relevant variables,
by country, as well as the Latin American average for comparison.
Across Central America, the reduction in fertility is generally lower than what is observed in the
Latin America region as a whole, which suggests that the demographic pressure on schooling
20
systems remained at relatively higher levels in the sub-region.23 However, the incorporation of
individuals belonging to households in the poorest 40 percent is lower in Central America, which
may have limited the extent of additional strain on education systems.24 The macroeconomic
context also seems to have provided a favorable environment in Central America, as was the case
for Latin America, although economic growth rates were below the regional average (with the
exception of Panama, where GDP growth almost doubled that of Latin America). Inflation rates
were also at or below the Latin America average, with only Costa Rica standing out with a higher
than average rate.
The performance of labor markets in Central America in the two dimensions considered is mixed,
since average wages increased at a lower rate than in Latin America, with declines in El Salvador
and Guatemala. The exception is Costa Rica, where average wages increased 11.5 percent.
However, employment rates across Central America increased relatively faster than the Latin
America average, with the exception of El Salvador. For the returns to education, a mixed picture
also emerges, with greater declines in the returns to higher education in Costa Rica, El Salvador,
Guatemala, and Panama relative to the Latin America average, but considerable increases in
Honduras and Nicaragua. The relative returns to US declined more than the Latin America average
in El Salvador, Nicaragua, and Panama, but increased considerably in Honduras.
Across Central America, then, the negative effects of demand variables are counterbalanced by the
positive effects of labor market variables, with marginal impacts from macroeconomic changes
and fluctuations in the returns to schooling. El Salvador stands out as an exception, where changes
across all variable groups would imply negative effects on enrollment. The results of this
decomposition of the change in secondary enrollment rates across cohorts for each country,
estimated by multiplying the changes in the independent variables by the coefficients from the
base regression, are summarized in Figure 18. Focusing specifically on labor market opportunities,
one very interesting piece of causal information comes from a natural experiment exploiting the
increase in demand of unskilled workers from the growth of factories in Mexico in the late 1980s
and 1990s. Atkin (2012) finds that this increase caused a rise in secondary dropout rates, as the
opportunity cost to staying in school rose.
Young people in Central American countries with high rates of out-migration are also likely to
consider not only their domestic labor market, but also opportunities in international labor markets
when weighing both the opportunity costs and expected returns to persisting in school. While
changes in international labor markets are not accounted for in the model, we look at a snapshot
of the education of migrants to the United States for suggestive evidence. In particular, we look
at the schooling participation and attainment of young people in the United States from the highest
23
Honduras and Nicaragua show greater reductions in fertility than the Latin America average.
24 Costa Rica has incorporated a greater share of students from the bottom 40 percent than the Latin America
average.
21
Central American migrant-sending countries (El Salvador, Guatemala, and Honduras), compared
to their counterparts at home. As shown in Table 7, over 90 percent of youth ages 12-17 are in
school, meaning that relatively few young migrants of LS and US ages are in America and
working. At the same time, young adults ages 18-25 who migrated to the United States at age 18
or later (so were likely to have completed any secondary schooling at home) have higher
educational attainment than their counterparts at home. These results suggest that migrants are
positively selected on education, and that labor market opportunities in the United States may not
be a major factor influencing dropout in Central America. However, population data in the United
States undercounts unauthorized migration, and further research would be needed to understand
the decision-making processes behind migration in order to provide a more thorough assessment.
Recently released summary results from the application of the TERCE (Third Regional
Comparative and Explanatory Study) exam in 2013 across Latin America show that Central
American students underperform the Latin America average, with the clear exception of Costa
Rica (UNESCO 2015). A first snapshot of summary results is presented in Figure 19, for all the
Central America countries participating (all except El Salvador). At both 3rd and 6th grades,
students in Nicaragua, Honduras, Guatemala, and Panama underperform the Latin America
average, and 30-40 percent are projected to fall into the lowest two performance levels (of five) in
each country (compared to 26 percent across Latin America). Costa Rica stands out as a strong
performer, with only 11 percent of students projected to fall into the lowest two levels (UNESCO
2015). Compared to SERCE (Second Regional Comparative and Explanatory Study) results from
2006, learning outcomes have improved in Guatemala, Panama, and to a lesser extent in
Nicaragua. Performance in Costa Rica has declined (Figure 20).
In a global context, education quality and learning outcomes in Central America are also low. In
Honduras, 6th grade students taking the 4th grade TIMSS 2011 (Trends in International
Mathematics and Science Study) mathematics assessment scored well below the international
22
average. In the 2012 PISA (Program for International Student Assessment), Costa Rica’s 15 year-
old students performed near the bottom of the international distribution. In math, Costa Rica
ranked 55 of 64 countries, and 51 of 64 in sciences. These results suggest that, while the quality
of primary education in Costa Rica (as measured in SERCE and TERCE) is relatively strong,
secondary education is weaker. They may also suggest that other Central American countries
would have even weaker outcomes at the secondary level on the global scale, given Costa Rica’s
outperformance in the regional exams.
These weak results corroborate negative perceptions held by citizens of the quality of public
education and the value of education in improving labor market outcomes across Central America,
as measured in the 2011 Latinobarometro surveys (World Bank 2014b and 2014c). For example,
in Honduras, citizens give their public education system one of the lowest rankings in Latin
America. At the same time, there is also substantial variation in quality within countries; as is the
case with access, outcomes for youth at the top of the socioeconomic distribution are much better
than for those at the bottom. For example, in Costa Rica, a 1 standard deviation increase in a
socioeconomic index correlates to a 0.3 standard deviation increase in SERCE scores (World Bank
2014d). Low average educational quality in many cases masks large disparities, whereby strong
results at the top of the socioeconomic distribution co-exist with very weak ones.
3.4 How dropout decisions are actually made, and evidence from qualitative
research
Taken together, human capital theory, the limited causal evidence available, and the empirical
correlations presented above show that a host of factors affect the marginal costs and benefits of
staying in school, and ultimately the decision to persist or drop out. From a public policy
perspective, this could be viewed as promising, in that there are many “levers to pull” or
opportunities for policy to impact on the dropout decision. At the same time, neither theory nor
empirics provide clear guidance on which factors may be the most important and therefore should
be the priorities for policy interventions.
In addition to the evidence already presented, recent developments in behavioral economics also
help shed light on the decision-making process that leads to dropout. As described in the 2015
World Development Report, individuals more often make decisions based on mental models,
drawing on information in their immediate social contexts and recent experiences, rather than a
careful weighing of all costs and benefits. Individuals also differ in what they even consider as
being choices within their own power to make. Some recent qualitative research on youth and
particularly on dropouts helps to highlight these factors and their potential policy implications.
Dropping out of school is often not a discrete decision, but the culmination of an ongoing process
of disengagement. Rich administrative data from the United States and other high-income
countries has shown that students who will eventually drop out prior to completing secondary
school can be reliably identified in primary school, based on their attendance, behavioral records,
23
and course performance (Gardner Center 2011; Frazelle and Nagel 2015). Interviews with youth
in Central America who have dropped out reflect these early origins of their ultimate decision. In
Costa Rica and Panama, many dropouts report having had trouble passing courses in the past and
sometimes repeating grades (Programa Estado de la Nacion 2013; UNICEF 2015). Students with
unaddressed learning deficits may get stuck at a grade level while their peers move on, or they
may be promoted to higher grades but unable to follow overly rapid curricula – in either case, they
are likely to eventually give up, or to be pushed out by school policies (Pritchett and Beatty 2015).
Poor or excluded youth may lack the agency to actively make the decision to drop out. In contexts
of poverty, individuals may suffer from excessive cognitive burdens in navigating the challenges
of daily life, leaving them little room for long-term planning (Mullainathan and Shafir 2013).
Living in conditions of poverty or social and economic exclusion also affects individuals’ attitudes
about themselves and their place in the world: they have lower aspirations and lower self-
confidence, reducing their propensity to deliberatively plan for the future (World Bank 2015c).
Interviews with adolescent girls in Nicaragua reflect this. Among the middle class, adolescent
girls articulate specific and concrete aspirations and plans for their own futures, and directly
connect educational achievement to achieving these plans. In contrast, poor girls are more focused
on short-term survival and, while generally believing in the value of education, do not apply it
directly to themselves (World Bank 2015d).
Adult support and guidance, or the lack of it, is an important factor in the dropout process for many
adolescents. In addition to the ways in which parents influence the human capital investments of
youth discussed in previous sections, adults more broadly can have important impacts. Dropouts
surveyed in Panama cited negative experiences with teachers, and many “dropped out” because
they were ultimately denied registration or expelled. In addition, students with similar academic
histories who persisted in school tend to describe having more support from parents or other adults
in their personal lives. Relatedly, nearly half of teachers surveyed in the same communities cite
students’ lack of motivation, poor academic performance, and lack of support at home as drivers
behind student dropout (UNICEF 2015). For youth who do not receive adequate support and
guidance at home nor in school, making poor decisions such as dropping out may be the “path of
least resistance” and the easiest choice to make in the short term.
24
These aspects of adolescent decision-making and dropout can help inform policy and program
interventions, as discussed in the next section.
A growing evidence base points to promising options for reducing dropout, but many challenges
remain for researchers and policy makers, including estimating costs, and improving targeting and
coordination across interventions to achieve the largest and most cost-effective impacts. In a
recent narrative review, Almeida, Fitzsimons, and Rogers (2015) assess the small but growing
body of global evidence from rigorously evaluated programs and policies to reduce dropout, with
a focus on secondary dropouts.25 In this section, we summarize their insights, and link them to the
Central American experience. Table 8 lists the main types of policies and programs that have been
implemented throughout the world that address underlying factors linked to the marginal benefits
and costs of dropout, broadly grouped by the level at which they act (individual/household versus
community or macro). Given the lifecycle nature of the dropout decision, we also attempt to
generally distinguish between interventions aimed primarily at preventing dropout by making
improvements that benefit entire student populations, or are aimed at ages prior to when substantial
dropout occurs, versus those aimed at intervening particularly for those youth who are at risk of
dropping out in the near future. In many cases, this distinction is not clear-cut. For example, broad
programs are often targeted in their implementation, due to budget constraints and need, and in
these targeted schools or communities, youth face many risks, including dropout.
In Central America, preschool coverage remains limited, particularly among the poor. In Costa
Rica, where there is no public spending on preschool, only 25 percent of 3-5 year olds in the bottom
quintile are enrolled, compared to 53 percent in the top quintile (World Bank 2014d). In Panama,
25
The authors define rigorous evaluations as those that employ a plausible counterfactual, including randomized
control trials, difference-in-difference, regression discontinuity, and other empirical designs.
26
All evidence described in this section that is not cited here is cited in detail in Almeida, Fitzsimons, and Rogers
(2015).
25
where an early childhood development plan was adopted in 2009 and is currently waiting
refinement, only 32 percent of 4 year-olds in the bottom quintile are in school, compared to 72
percent in the top quintile (World Bank 2014c). At the same time, while early intervention
programs targeting different risk groups may have been tried across countries, we were not able to
identify any that had been rigorously evaluated or brought to national scale.
4.2 Household wealth and individual motivation: Providing financial and non-
financial incentives
Given that economic factors are one of the main reasons cited by dropouts in Central America,
programs and policies to counteract the financial opportunity costs of remaining in school and
increase the immediate benefits may be particular important. Programs that have been rigorously
evaluated around the world generally find substantial and significant positive impacts on school
participation, attendance, and persistence at the primary and lower secondary levels. At the
upper secondary level, evaluations of transfer and scholarship programs find that the existence of
other support programs (reducing the marginal benefit of additional cash) and the timing of
transfers may be particularly important for determining impacts. The impacts of changing the
timing of transfers in particular reflects in part the value of addressing individuals’ time
inconsistent preferences, by facilitating savings for them to pay for up-front costs associated with
re-enrolling (Barrera-Osorio et al 2011).
Every Central American country except Nicaragua currently has a conditional cash transfer
program which includes education conditionalities. In El Salvador, Guatemala, and Honduras,
the programs’ education conditionalities cover only youth up to either 15 years old or 6th grade
(Cecchini and Madariaga 2011). In Panama, Red de Oportunidades includes youth up to 17,
while Beca Universal includes all students in public schools and private schools charging fees
below a certain amount. In Costa Rica Avancemos covers all youth attending secondary school.
Targeting, coverage, and verification of conditionalities are major challenges that limit the
impacts of some programs. Panama and Costa Rica’s universal scholarship programs (Beca and
Avancemos) are, by their nature, not well-targeted. In Panama, the majority of Beca
beneficiaries are in the middle income deciles (3rd to 9th) rather than among the poorest, while in
Costa Rica, only 29 percent of eligible students in the bottom quintile reporting receiving
Avancemos compared to 8 percent in the top quintile (World Bank 2014c and 2014d). These
programs have not been rigorously evaluated, and their impacts on school outcomes are not
clear. Even well-targeted programs with strong evaluation results, like El Salvador’s
Comunidades Solidarias Rurales do not cover all or most eligible households: in El Salvador,
fewer than 12 percent of households in the poorest quintile are beneficiaries (World Bank
2014e). Other programs providing materials, food, and other inputs to school could also be
better targeted in some cases. For example, only 24 percent of students in the bottom quintile in
Costa Rica benefit from the school meal program, compared to 12 percent in the fourth and 6
percent in the fifth.
26
Other types of incentives, such as performance-based financial rewards for students and school
vouchers, have shown some promise around the world. Evaluations of merit scholarships and
achievement-based financial rewards have found mixed results, and Almeida, Fitzsimons, and
Rogers suggest that these interventions may be successful in increasing learning and reducing
dropout at the secondary level in contexts where other constraints, such as motivation and
teacher quality, are not binding and where other financial incentives are not already in place.
Similarly, school vouchers appear to benefit students who are motivated enough to apply, in
contexts where voucher-accepting schools are available and of significantly higher quality than
the alternatives. This may be an approach to consider in urban settings in Central America,
where the market for schools is sufficiently “thick” to give students options.
27
On the intensive margin, students already enrolled may spend more time in school through policies
that extend the school day. As fertility rates have fallen and reduced pressure on school
infrastructure and personnel, several Latin America countries have implemented a form of school
day extension, including El Salvador’s full-time school program. A recent review of such
extension programs finds that the majority of impacts measured in rigorous evaluations are positive
and significant, and particularly substantial for at-risk students (Holland, Evans, and Alfaro 2014).
The evaluations that look at persistence in school find significant increases in graduation rates.
A second area for broad quality improvements is teacher policy. Bruns and Luque (2014) provide
a comprehensive overview of how teacher recruitment, training, and incentive policies fail to
maximize teacher performance across Latin America. In Honduras, as in several other countries
in the region, even basic indicators such as class time used for actual instruction show substantial
opportunity for improvement – the equivalent of one day of school per week is lost to teachers
socializing or conducting management activities during class rather than teaching. A growing
number of randomized evaluations find that performance-based monetary incentives for teachers
can increase student learning, and presumably reduce dropout, under the right circumstances.
Almeida, Fitzsimons, and Rogers identify a pre-program lack of accountability and low levels of
teacher professionalism, coupled with relatively large incentives based on clear, recent, and fair
metrics as some of the needed circumstances.
In addition to system-wide efforts to improve quality, several countries have addressed the quality
of education provided to specific sub-groups. In particular, in Central America, bilingual and
culturally adapted education for indigenous peoples has shown positive impacts on learning, grade
28
progression, and ultimate educational attainment (Dutcher and Tucker 1994). Patrinos and Velez
(2009) show, for example, that bilingual education in Guatemala increases learning and reduces
dropout rates, and provides a substantial net benefit to the system.
Many different types of programs directly target youth considered at risk of dropping out to try
and build their cognitive and socio-emotional skills. As discussed earlier, since dropout is the
culmination of a process, identifying students who are most at-risk may be possible early on, if
accurate and detailed administrative data is available. To date, however, only one rigorous
evaluation has been conducted on the impacts of implementing such “early warning systems”,
along with targeted interventions, on dropout rates – the USAID-funded School Dropout
Prevention Pilot Program, whose initial results show limited effects on dropout across the four
countries involved in the study (Cambodia, India, Tajikistan, and Timor-Leste). Among students
lagging behind in terms of learning, individual academic tutoring has been shown to improve
learning at the primary level in India. On the other hand, a range of programs have used
information and computer technology to supplement instruction at the primary and secondary
levels – the bulk of the evidence finds that these programs have little to no effect on learning, and
therefore probably not on dropout either, because the technology is not well-utilized for instruction
and learning.
29
4.7 Addressing key risk factors in the community
Teenage pregnancy is correlated with dropout, as discussed earlier, but the causal pathways are
complex and multidirectional. The average adolescent fertility rate across Latin America is on par
with South Asia and far above all regions except Sub-Saharan Africa, and within Latin America,
Nicaragua has the highest rate, while Guatemala, Honduras, El Salvador, and Panama are all in the
top 10 (Azevedo et al 2012). Given the many negative consequences of early motherhood,
reducing teenage pregnancy is an important goal itself, but Almeida, Fitzsimons, and Rogers find
that there is limited rigorous evidence on programs directly aimed at reducing pregnancy, and the
evidence that does exist does not find significant impacts. One interesting natural experiment
exploited the expansion of family planning centers in Colombia, finding that the availability of
such services did increase young women’s educational attainment (Miller 2010). At the same
time, a review of programs aimed at keeping teen mothers in school finds that they are successful
in reducing dropout among these young women (Steinka-Fry, Wilson, and Tanner-Smith 2013).
Crime and violence are also correlated with dropout, and may increase direct risks faced by
students every day, the opportunity costs of staying in school, or other aspects of the dropout
decision. As discussed above, evidence from several countries supports the idea that increased
schooling has a causal impact on crime in the short and long run. In comparison, less is known
about whether crime reduction can independently reduce dropouts – in some sense, this may be a
more difficult question to answer convincingly, as natural experiments tend to provide temporal
variation in crime and violence, while dropout rates may only change slowly over time as they are
the outcome of a longer-term process. There is growing evidence that crime and violence have a
causal impact on student attendance, which can lead to lower attainment and dropout, and also that
large-scale violent conflict substantially reduces educational attainment for affected youth (Smith
2003; Bruck, Di Maio, and Miaari 2014).
30
While the knowledge base on what works is growing, much less is known about which approaches
to reducing dropout are most cost-effective. This is hugely important for governments in Central
America, where education spending relative to GDP is generally low, and resources available to
target this particular group are limited (World Bank 2014b, 2014c, 2014d, and 2014e). Some
programs have been costed and are known to be relatively inexpensive with light implementation
requirements, such as information campaigns, while others are much more costly and demanding
in terms of implementation, such as extending school hours (Holland, Evans, and Alfaro 2014).
However, we are still far from having systematically comparable cost data across different types
of programs and contexts, which would facilitate decision-making for policy makers (Evans and
Popova 2014).
Across all six Central American countries, there are common policy priorities which would likely
contribute to reducing dropout, including improving the quality of education at all levels,
addressing remaining gaps in access at the preprimary and secondary levels, and strengthening the
coverage, targeting, and coherence of existing programs aimed at improving education outcomes.
Beyond this broad set, each country’s particular circumstances can inform specific priorities –
initial ideas are discussed below in the concluding section.
5. Conclusions
While substantial progress in educational attainment has been made in Central America over the
last several decades, dropout remains a major challenge to building the skills of the workforce and
human capital of the population. As detailed in Section 1, each country has a unique history of
education progress, but today the sub-region faces some similar dropout problems. In Panama and
Costa Rica, most youth remain in school through US ages, but many are over age for their grade,
and most drop out before completing US. In Honduras, Guatemala, and Nicaragua, only a minority
of youth will reach US, with most dropping out at earlier stages. In El Salvador, the dropout profile
of the most recent cohorts is in between these two descriptions.
Within each country, large inequalities in dropout rates and therefore educational attainment
outcomes are evident. Poverty, rurality, and indigenous group membership are the strongest
correlates of dropout, likely reflecting several underlying factors that affect both the marginal
benefits and costs of staying in school. On the benefits side, these factors influence human capital
built early in life, the quality of schooling services available, and labor market opportunities. They
also affect the direct and opportunity costs of persisting, including the actual accessibility of
schools.
The consequences of dropout are felt at the individual and aggregate levels. As described in
Section 2, youth with less education have worse labor market outcomes, in terms of employment,
informality, and wages. Evidence from around the world also finds that educational attainment is
31
correlated with many other life outcomes, including health and the well-being of the next
generation. In aggregate, workforces that lack the skills demanded in the labor market constrain
growth at all stages of development, as has been pointed out in the Systematic Country Diagnostics
of Costa Rica, El Salvador, and Panama (World Bank 2014, 2015a, and 2015b). Youth lacking
skills and labor market opportunities are also likely to be more vulnerable to becoming involved
with crime or getting pregnant early.
Substantial progress has been made in understanding the drivers of dropout and identifying
interventions to reduce it. A range of factors influence the process that leads to eventual dropout,
and most of these factors can be affected by policy, as discussed in Section 3. Despite recent
growth in the knowledge base of what works to reduce dropout, the challenge of detailed
prioritization remains. As described in Section 4, the cost-effectiveness of different interventions,
as well as potential interactions across interventions, are not well-understood. In addition, the
design details of individual interventions may matter a great deal for impacts. Despite these
limitations, several conclusions can be drawn for Central America.
Across all six countries, there are common priority policy areas which would contribute to
reducing dropout, with different specifics depending on each country’s context and dropout
challenge. To prevent dropout from occurring in the long-term, building human capital from early
childhood onwards is critical, and requires coordination across sectors to address health, nutrition,
and parenting in the years before children enter the education system. Specific approaches will
vary by country, but could include targeting the most disadvantaged communities where early life
outcomes are known to be weakest. Within school systems, both de jure and de facto access are
in need of attention. In Nicaragua, making LS compulsory and expanding supply accordingly are
necessary first steps, while in Honduras, El Salvador, and possibly Guatemala, expanding supply
is needed to fulfill existing compulsory schooling laws. In Panama and Costa Rica, where US
participation rates are highest, the focus may need to be more on practices within schools, in terms
of increasing support for poorly performing students and changing school cultures to make
universal completion the expectation.
The quality of education provided, on average and in particular to lower socioeconomic groups, is
also a priority area which would contribute to preventing dropout. While learning outcomes at the
primary level in Costa Rica are relatively strong on average, large differences across
socioeconomic groups persist, and learning at the secondary level appears to be much weaker. In
all other Central American countries, average learning outcomes at the primary level are low, and
a host of promising reforms to increase autonomy, assessment, and accountability are at various
stages of implementation in several countries. To the extent possible, rigorous evaluations of these
efforts, such as the ones of teacher and principal trainings in El Salvador and Mexico, should be
conducted and should include cost data to help inform policy making in the region and beyond.
32
To directly intervene among youth at high risk of dropping out, countries can consider developing
early warning systems, and piloting comprehensive support programs for identified youth. Such
an approach would be most promising in countries with strong administrative data, and would need
to be targeted to the levels where dropout starts to occur at substantial rates. Given that dropout
rates are so high, school or community-level programs might make more sense than individual-
level programs.
Reforming the targeting, coverage, and coherence of existing programs would also help. As
discussed above, some conditional cash transfer and other support programs are not well-targeted,
while others are well-targeted but are too small to cover the majority of the eligible. In addition,
there may be opportunities to improve the design of existing programs to increase their impacts,
based on insights from behavioral economics, for example by structuring the timing of transfers to
maximize impacts on the decision to persist in school. Given that most countries have multiple
social protection and school support programs aimed at overlapping populations, rationalizing and
coordinating across programs may provide both increased impact and cost savings. Finally, based
on the evidence that child labor is commonly used as a coping strategy in response to shocks in
the region, rapid scale-up mechanisms for existing programs, in terms of expanding coverage or
increasing support following shocks, could also play a role in reducing dropout.
33
Bibliography
Acosta, Pablo. 2011a. “Female Migration and Child Occupation in Rural El Salvador.”
Population Research and Policy Review 30 (4).
Acosta, Pablo. 2011b. “School Attendance, Child Labour, and Remittances from International
Migration in El Salvador.” Journal of Development Studies 47 (6): 913-936.
Ambler, K., D. Aycinena and D. Yang. 2014. “Channeling Remittances to Education: A Field
Experiment Among Migrants from El Salvador.” NBER Working Paper 20262. NBER:
Cambridge, MA.
Atkin, David. 2012. “Endogeneous Skill Acquisition and Export Manufacturing in Mexico.”
NBER Working Paper 18266. NBER: Cambridge, MA.
Avitabile, Ciro and Rafael de Hoyos. 2014. “Heterogeneous Effects of Information about the
Returns to Schooling on Student Learning: Evidence from a Randomized Control Trial in
Mexico.” World Bank Working Paper. World Bank: Washington, D.C.
Azevedo, Joao Pedro, Marta Favara, Sarah Haddock, Luis Lopez-Calva, Miriam Muller, and
Elizaveta Perova. 2012. Teenage Pregnancy and Opportunities in Latin American and the
Caribbean. World Bank: Washington, D.C.
Baez, Javier and Indhira Santos. 2007. “Children’s Vulnerability to Weather Shocks: Natural
Disasters as a Natural Experiment.” Unpublished Working Paper.
Baez, Javier and Indhira Santos. 2008. “On Shaky Ground: The Effects of Earthquakes on
Household Income and Poverty.” United Nations Development Program: New York, NY.
Baez, Javier, Alejandro de la Fuente, and Indhira Santos. 2010. “Do Natural Disasters Affect
Human Capital? An Assessment Based on Existing Empirical Evidence.” IZA Working Paper
5164.
Baez, Javier, Leonardo Lucchetti, Maria Genoni, and Mateo Salazar. 2015. “Gone with the
Storm: Rainfall Shocks and Household Wellbeing in Guatemala.” World Bank Working Paper.
World Bank: Washington, D.C
Barro, Robert and Jong-Wha Lee. 2010. “A New Dataset of Educational Attainment in the
World, 1950-2010.” Journal of Development Economics 104: 184-198.
Bassi, M., M. Busso, Sergio Urzúa and J. S. Muñoz. 2013 “Is the Glass Half Empty of Half Full?
School Enrollment, Graduation and Dropout Rates in Latin America.” IDB Working Paper
Series No. IDB-WP-492. Inter-American Development Bank: Washington, D.C.
Bentaouet-Kattan, R., and Székely, M. 2015b. “Patterns, Consequences and Possible Causes of
Dropout in Upper Secondary Education in Mexico.” Education Research International, Volume
2015, Article ID 676472.
Bentaouet-Kattan, R., and Székely, M. 2015. “Analyzing the Dynamics of Upper Secondary
Schooling in Latin America: A Cohort Approach.” World Bank Working Paper. World Bank:
Washington, D.C.
34
Branson, N., C. Hofmeyr, and D. Lam. 2014. “Progress through school and the determinants of
school dropout in South Africa.” Development Southern Africa (31) 1: 106-126.
Bruck, Tilman, Michele Di Maio, and Sami Miaari. 2014. “Learning the Hard Way: The Effect
of Violent Conflict on Student Academic Achievement.” Households in Conflict Network
Working Paper 185. University of Sussex: Brighton, England.
Bruns, Barbara and Javier Luque. 2015. Great Teachers: How to Raise Student Learning in
Latin America and the Caribbean. World Bank: Washington, D.C.
Bustelo, Monserrat. 2015. “Bearing the Burden of Natural Disasters: Child Labor and
Schooling in the Aftermath of Tropic Storm Stan in Guatemala.” World Bank Working Paper.
World Bank: Washington, D.C
Cárdenas, M., de Hoyos, R., y Székely M. “Idle Youth in Latin America: a persistent problem in
a decade of prosperity.” Economia, forthcoming, 2015.
Cardoso, A.R., and D. Vermer. 2007. “School drop-out and push-out factors in Brazil: The role
of early parenthood, child labor, and poverty.” World Bank Working Paper WPS4178. World
Bank: Washington, D.C.
Cecchini, Simone and Aldo Madariaga. 2011. Conditional Cash Transfer Programmes: The
Recent Experience in Latin America and the Caribbean. UN ECLAC: Santiago, Chile.
Chioda, Laura. 2013. Crime and Violence over the Life cycle in Latin America and the
Caribbean. World Bank: Washington, D.C. Cox Edwards, A., and M. Ureta. 2003.
“International migration, remittances, and schooling: evidence from El Salvador.” Journal of
Development Economics 72: 429– 461.
Cunha, Flavio and James Heckman. 2007. “The Technology of Skill Formation.” American
Economic Review 97(2): 31-47.
Cunningham, Wendy, Linda McGinnis, Rodrigo García Verdú, Cornelia Tesliuc, and Dorte
Verner. 2008. Youth At Risk in Latin America and the Caribbean: Understanding the Causes,
Realizing the Potential. World Bank: Washington, D.C.
Dearden, L., C. Emmerson, C. Frayne, and C. Meghir. 2008. “Conditional Cash Transfers and
School Dropout Rates.” Journal of Human Resources 44 (4).
Dutcher, N. and G. Tucker. 1994. “The Use of First and Second Languages in Education.”
Working paper. World Bank: Washington, D.C.
Estado de la Nación. 2013. Estado de la Educación Costarricense. Estado de la Nación: San
José, Costa Rica.
Ferreira, Francisco, Julian Messina, Jamele Rigolini, Luis-Felipe Lopez-Calva, Maria Ana Lugo,
and Renos Vakis. 2012. Economic Mobility and the Rise of the Latin American Middle Class.
World Bank: Washington, D.C.
Fiszbein, Ariel, Schady, Norbert Rüdiger, & Ferreira, Francisco HG. 2009. Conditional cash
transfers: reducing present and future poverty. World Bank: Washington, D.C.
Frazelle, Sarah and Aisling Nagel. 2015. A Practitioner’s Guide to Implementing Early
Warning Systems.
35
Fryer, Roland. 2013. “Information and Student Achievement: Evidence from a Cellular Phone
Experiment.” NBER Working Paper 19113.
Gardner Center. 2011. Using Early Warning Systems to Predict and Prevent Dropout: Youth
Data Archive Policy Factsheet. Stanford University: Palo Alto, CA.
Hanushek, Eric, Victor Lavy, and Kohtaro Hitomi. 2008. “Do Students Care About Quality?
Detreminants of Dropout Behavior in Developing Countries.” Journal of Human Capital 2(1):
69-105.
Ibáñez, A.M, C. Rodríguez, and D. Zarruk. 2013. “Crime, Punishment, and Schooling Decisions:
Evidence from Colombian Adolescents.” IDB Working Paper IDB-WP-413. Inter-American
Development Bank: Washington, D.C.
Ingrum, A. 2006. “High School Dropout Determinants: The Effect of Poverty and Learning
Disabilities.” The Park Place Economist, Volume XIV.
Jensen, Robert. 2010. “The (Perceived) Returns to Education and the Demand for Schooling.”
The Quarterly Journal of Economics 125 (2): 515-548.
Jimenez, W. and M. Gaete. 2010. “Abandono (deserción) escolar en la enseñanza secundaria en
Costa Rica, 2009-2010.” Ministerio de Educación Pública, Costa Rica.
Manacorda, M. 2006. “Grade Failure, Drop out and Subsequent School Outcomes: Quasi-
Experimental Evidence from Uruguayan Administrative Data”, Mimeo, STICERD, London,
Marcotte, D. 2013. “High school dropout and teen childbearing.” Economics of Education
Review 34: 258–268.
Marshall, J.C. Aguilar, M. Alas, R. Castellanos, L. Castro, R. Enamorado, and E. Fonseca. 2014.
“Alternative Education Programmes and Middle School Dropout in Honduras.” Int Rev
Education 60:51–77.
Ministerio de Educación de Nicaragua. 2014. “Reporte de Implementación hasta 2013, Plan
Estratégico de Educación (PEE) 2011-2015.” Ministerio de Educación, Managua.
Ministerio de Educación de Guatemala. 2014. “Informe de revisión nacional de la educación
para todos, Guatemala 2000 – 2013.” Guatemala.
Mo, D., L. Zhang, H. Yi, R. Luo, S. Rozelle, and C. Brinton, 2014. “School Dropouts and
Conditional Cash Transfers: Evidence from a Randomised Controlled Trial in Rural China's
Junior High Schools.” Journal of Development Studies 49 (2): 190-207.
Mullainathan, Sendhil and Eldar Shafir. Scarcity: Why Having Too Little Means So Much.
Picador.
Munyo, I. 2013. “Youth Crime in Latin America: Key Determinants and Effective Public Policy
Responses.” The Brookings Institution.
Nguyen, Trang. 2008. “Information, Role Models, and the Perceived Returns to Education.”
Unpublished manuscript.
O’Donoghue, Ted and Matthew Rabin. 2001. “Risky Behavior among Youths: Some Issues
from Behavioral Economics.” In Risky Behavior among Youths: An Economic Analysis, editor
Jonathan Gruber. University of Chicago Press: Chicago, IL.
36
OEI. 2010. “2021 Metas Educativas: La Educación Que Queremos para la Generación de los
Bicentenarios.” Organización de Estados Iberoamericanos para la Educación: Madrid, Spain.
Oreopoulos, Philip and Kjell Salvanes. 2011. “Priceless: The Nonpecuniary Benefits of
Schooling.” Journal of Economic Perspectives 25(1): 159-184.
Patrinos, Harry and Eduardo Velez. 2009. “Costs and benefits of Bilingual Education in
Guatemala: A Partial Analysis.” International Journal of Education Development 29: 594-598.
37
World Bank. 2014f. Romania: Draft Strategic Framework to Reduce Early School Leaving.
World Bank: Washington, D.C.
World Bank. 2015a. Costa Rica Systematic Country Diagnostic. World Bank: Washington,
D.C.
World Bank. 2015b. El Salvador Systematic Country Diagnostic: Building on Strengths for a
New Generation. World Bank: Washington, D.C.
World Bank. 2015c. World Development Report 2015: Mind, Society, and Behavior. World
Bank: Washington, D.C.
World Bank. 2015d. Qualitative Study on Teenage Motherhood in Nicaragua: Overview. World
Bank: Washington, D.C.
38
Figure 1
Note: Structural peers are defined in each country’s Systematic Country Diagnostic. Costa Rica: Chile, Croatia,
Dominican Republic, Lithuania, Panama, Uruguay. Panama: Bulgaria, Costa Rica, Croatia, Dominican Republic,
Uruguay. El Salvador: Armenia, Bosnia and Herzegovina, Dominican Republic, Georgia, Serbia, and Tunisia.
Source: Barro and Lee 2010
Figure 2
39
Figure 3a
Figure 3b
40
Figure 4a
Figure 4b
41
Figure 5
Figure 6
42
Figure 7
Figure 8
43
Figure 9
Figure 10
44
Figure 11
Figure 12
45
Figure 13
Figure 14
46
Figure 15
12-17
Figure 16
12-17
47
Figure 17
12-17
Figure 18
48
Figure 19
Figure 20
49
Diagram 1
Pregnancy………………………………………………..
Crime and Violence………………………………….
Lack of access: limited seats, distance…………….
Labor market opportunities (e.g. migration, maquilas)
Financial barriers: fees and costs related to attendance……….
marginal costs
Indicators of
secondary
drop out?
Primary
Early
Irrelevant/unappealing content………………………..
Childhood
Low perceptions or lack of information on returns
marginal benefits
50
Table 1
Table 2
51
Table 3
Table 4
Source: Author´s calculations with household survey data. Schooling gap is calculated as the value of
expected years of formal education for a given age, and the actual years attained
52
Table 5
Personal characteristics
-
Respondent is female 0.040 * -0.038 * -0.041 *
-
Respondent is indigenous 0.050 * -0.045 * -0.048 *
Household characteristics
Years of schooling HH head 0.022 * 0.011 * 0.01 *
Household Size 0.002 -0.001 -0.001
-
Single parent household 0.029 * -0.025 * -0.028 *
-
Respondent lives in rural area 0.132 * -0.109 * -0.075 *
53
Table 6
Change in the value of independent variables explaining school dropout during 2000-2012
Macro
Demand side variables
variables Labor market variables Change in returns to schooling
Country Change in Change in the % Annual (Ages 25-45) (Ages 25-55)
Growth
Fertility of poor
of Inflation Change in Change in Higher Educ. US
Rate youth
across graduating GDP per Rate Real Average Employment Relative to Relative to
Cohorts from LS capita Wages Rate Primary Primary
Costa Rica -1.33 15% 2.6% 15.0% 11.50% 7.0% -19.6% -2.0%
El
12%
Salvador -1.69 1.4% 4.0% -6.40% 3.0% -12.5% -30.7%
Guatemala -1.61 10% 0.8% 8.0% -1.90% 8.0% -19.7% -1.1%
Honduras -1.99 7% 2.1% 10.0% 0.00% 11.0% 13.6% 17.4%
Nicaragua -2.12 8% 1.7% 11.0% 8.20% 11.0% 6.1% -18.5%
Panama -0.53 6% 4.9% 3.0% 3.50% 12.0% -9.7% -8.9%
Total
Latin 12%
America -1.90 2.5% 11.5% 9.30% 6.0% -3.5% -6.7%
Source: Betauet-Kattan and Székely (2015).
54
Table 7
55
Table 8
Adapted from Almeida, Fitzsimons, and Rogers (2015) and World Bank (2014f )
56
Appendix
Table A.1
El Salvador 1989 1992 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2007 2008 2009 2010 2011 2012 2013 19
Panamá 1991 1995 1996 1997 1998 1999 2001 2002 2003 2004 2006 2007 2008 2009 2010 2012 2013 17
Honduras 1989 1992 1996 1997 1998 1999 2001 2002 2003 2005 2007 2008 2009 2010 2011 2012 16
Costa Rica 1987 1989 1991 1993 1995 1997 1998 2000 2001 2002 2003 2004 2009 2010 2012 2013 16
Guatemala 1998 2000 2004 2006 2009 2010 2011 2012 2013 9
57
Table A.2
Table A.3
58
Table A.4
Personal characteristics
Respondent is female 0,053 - 0,040 - 0,080 0,056 0,028 0,033
0,019 0,014 0,035 0,023 0,019 0,027
Household characteristics
Years of schooling HH head 0,013 0,022 0,045 0,032 0,027 0,012
0,008 0,013 0,023 0,019 0,016 0,008
Respondent lives in rural area - 0,019 - 0,132 - 0,149 - 0,160 - 0,144 - 0,065
0,011 0,041 0,041 0,044 0,034 0,015
Source: Author´s calculations by using household survey data. The coefficients for marginal effects
derived from a Probit estimation, are presented. Standard errors are shown in italics below the marginal
effects.
59
Table A.5
Country Average value of Coefficient for Primary Average value of Coefficient for Secondary Average value of Coefficient for Higher Education
1990s 2000s 2010s 1990s 2000s 2010s 1990s 2000s 2010s
Costa Rica 0,16 0,29 0,21 0,31 0,59 0,41 0,46 0,93 0,53
El Salvador 0,66 0,41 0,35 1,07 0,70 0,70 1,37 0,89 0,99
Guatemala 0,35 0,43 0,79 0,68 1,93 1,06
Honduras 0,25 0,22 0,68 0,51 0,38 1,19 0,85 0,63 1,60
Nicaragua 1,23 0,81 0,45 2,25 1,51 0,86 3,01 2,08 1,16
Panamá 0,13 0,23 0,34 0,29 0,52 0,62 0,48 0,80 0,88
Note: The above table contains the average value of the coefficient of the standard Mincer regression where the
dependent variable is the log hourly wage of each worker, and the independent variables are dummies for Primary,
Secondary (including lower and upper), and Higher education. The reference variable is no schooling, so the
coefficient reflects the additional income that is obtained when having certain education level, with respect to having
no schooling. The standard Heckman correction was used to account for labor participation potential bias in the
regression. The figures reported in the text refer to the relative returns between two different levels (the Secondary
over the Primary coefficient and the Higher Education over the Primary coefficient). Average coefficients are included
for brevity (which are the coefficients actually used to calculate the relative returns mentioned in the text), since the
full estimations include one regression per country/year – that is, 84 different regressions. In these base regressions it
is observed that all coefficients are statistically significant.
60
Figure A.1
Figure A.2
61
Figure A.3
Figure A.4
62