Studies in Educational Evaluation 58 (2018) 80–96

Contents lists available at ScienceDirect

Studies in Educational Evaluation

journal homepage: www.elsevier.com/locate/stueduc

The effect of districts’ social development on student performance☆ T

⁎
Gregorio Gimenez , Ángel Martín-Oro, Jaime Sanaú
Universidad de Zaragoza, Gran Vía, 2, 50005 Zaragoza, Spain

A R T I C LE I N FO A B S T R A C T

Keywords: This paper explores the relationship between student performance and social development in the districts where
Education schools are located. The empirical research, which focuses on Costa Rica, combines geolocalized data of schools
Social development from PISA 2012 with a composite index of social development that has four components: health, economy,
Districts education and electoral participation. Our results show that social development has positive, but diminishing
Schools
effects, on test scores and accounts for 11.6%, 13.1% and 14.8%, respectively, of the total variance in math,
Student performance
Costa Rica
reading and science. The economic and education dimensions of social development positively correlate with
PISA academic achievement. In the case of electoral participation, the estimations show a significant effect only in
math. We find no significant effect of health. Educational policies should give particular emphasis to the lowest-
ranked districts where the expected returns from policies that increase social development are the highest.

1. Introduction
There is a growing interest in explaining for the determinants in
educational achievement as an important factor behind economic de-
velopment (Hanushek & Woessmann, 2015). The underperformance of
Latin American education systems and the intensity of the skills gap in
this region compared to other emerging regions (Hanushek &
Woessmann, 2012; Melguizo & Perea, 2016) motivated us to focus on
Latin America and to explore the relationship between educational
achievement and the socio-economic characteristics of the school dis-
tricts. On this continent, there are large student performance gaps
within the same town or city and even between schools located very
close together, a reflection of the high-income inequalities in Latin
American urban areas.
We focus on the interaction between educational achievement and
social development at district level. This relationship has been rela-
tively unexplored because of data availability problems. We handle this
issue in our research by focusing on Costa Rica, which allows us to work
with highly disaggregated data. Our empirical analysis requires
homogeneous information with a high degree of geographical dis-
aggregation for a large set of variables that include educational and
social data. Costa Rica is one of the countries in the region with more
accurate statistics, both with regards to the variety of the information
collected and in the methodological quality of its elaboration. In
addition, this country brings together a set of characteristics that are
very relevant for this study. Costa Rica is ranked as a country of high
human development, according to United Nations. With a value of
0.776 in the 2015 Human Development Index, it is ranked 66th in the
world and 5th in Latin America. Its life expectancy is the highest of the
region (79.6 years) and it has one of the oldest and most stable
democracies on the continent.
The hypothesis to be tested is whether social development—defined
in a multidimensional perspective encompassing health, economy,
education and electoral participation—has a significant effect on aca-
demic achievement. Our methodological approach provides several
contributions to previous research: 1) the combination of the OECD
PISA 2012 dataset and a district-level Social Development Index (SDI);
2) the use of educational data with three-level hierarchical models
(students, schools and districts), which allows us to deal with selection
bias and unobserved heterogeneity; 3) the analysis of non-linearities;
and 4) the use of the Shapley-Shorrocks decomposition of the variance
to estimate the relative contribution of the SDI to the variance of the
test scores. The empirical work suggests that social development has a
positive, but diminishing, effect on test scores and explains a significant
part of their variance. This result is robust to the different disciplines
(math, reading and science).
The rest of the paper is organized as follows. Section 2 gives an
overview of the literature on the determinants of educational

☆
We wish to express our gratitude to the Estado de la Educación program in Costa Rica for their technical support and valuable ideas, especially to Isabel Román, Jennyfer León and
Eiliana Montero. Also, to Olegario Sáenz, from the Ministerio de Planificación Nacional y Política Económica of Costa Rica, for his help with the data. We also thank the anonymous
referees for their helpful comments and suggestions. This work was supported by the Fondo Concursable del Estado de la Educación of CONARE (Costa Rica) and the European Regional
Development Fund (operational program Building Europe from Aragon) – Department of Innovation, Research and University of the Government of Aragon (S36).
⁎
Corresponding author.
E-mail addresses: gregim@unizar.es (G. Gimenez), amartinoro@inbestia.com (Á. Martín-Oro), jsanau@unizar.es (J. Sanaú).

https://doi.org/10.1016/j.stueduc.2018.05.009
Received 2 October 2017; Received in revised form 27 May 2018; Accepted 29 May 2018
0191-491X/ © 2018 Elsevier Ltd. All rights reserved.
G. Gimenez et al.
achievement, highlighting in different dimensions of social develop-
ment. In Section 3, the data and empirical methodology employed in
the paper are presented. Section 4 shows the results of the estimation of
the education production function. Section 5 contains the Shapley-
Shorrocks analysis. Section 6 provides a summary and conclusions.
2. Literature background
Research has shown that not only school and family, but also the
“wider environment” in which students interact affect their outcomes.
Indeed, as Rasbash, Leckie, Pillinger, and Jenkins (2010, p. 657) state
that, “children are raised in complex social environments that involve
multiple layers of influence”, following Bronfenbrenner (1977). As well
as the more conventional family and school factors, local area influ-
ences may also be worth considering. One strand of multidisciplinary
literature deals with the effects that living in a certain area may have on
a series of outcomes related to physical and mental health, future
earnings, behavioral problems in youngsters, or educational achieve-
ment. Different types of causal mechanisms, such as community net-
works or role models, have been put forward by researchers (Burdick-
Will et al., 2011; Durlauf, 2004; Galster, 2012; Sampson, Morenoff, &
Gannon-Rowley, 2002; Van Ham & Manley, 2012; Vigdor & Ludwig,
2010). In this section, the literature linking student success and social
development is reviewed. We focus on four dimensions: health,
economy, education and electoral participation.
2.1. Health
A local environment with better health conditions, such as birth
weight, malnutrition or access to safe water, directly or indirectly in-
fluences students’ cognitive abilities. A healthier environment posi-
tively influences children’s ability to learn and, hence, acquire skills
(Case, Fertig, & Paxson, 2005; Currie & Goodman, 2010; Miguel &
Kremer, 2004; Ozier, 2014). For instance, sight or hearing difficulties,
which hinder school performance, are more likely to affect children in
lower-income families (Rothstein, 2010). Poor health indicators in a
community may signal poor socioeconomic conditions, which do not
provide children with a supportive environment for academic
achievement.
Nevertheless, some empirical studies in developing countries have
questioned this relationship (Aguilera & Quintana, 2011; Evans &
Popova, 2015) or conclude that health programs can be ineffective
(Conn, 2014; McEwan, 2015).
2.2. Economy
Studies show that children who live in extreme poverty or who lack
basic needs like food, clothing, shelter or infrastructure—such as elec-
tricity or access to the Internet— are more likely to perform poorly at
school and are also very likely to drop out. Dual factors of socialization
and social status have an influence on the achievement of students who
attend school (Harris, 2006; Jensen, 2009).
In the US context, the evidence from the Moving to Opportunity
experiment of the 1990 s is worth to be considered.1 Recent research
shows the importance of neighborhood economic conditions for chil-
dren’s upward economic and social mobility, as moving to lower-pov-
erty areas at a very young age has significant benefits (Chetty, Hendren,
& Katz, 2016). In a different context for Milwaukee (Wisconsin),
Carlson and Cowen (2015) study the relationship between the socio-
economic characteristics of the neighborhood and student outcomes.
They find evidence of the importance of neighborhoods as a
1
This unique policy experiment, designed and implemented by the U.S. Department of
Housing and Urban Development in the 1990s, aimed to shed light on whether providing
assistance to low-income families to move to better neighborhoods could improve their
outcomes.
81
Studies in Educational Evaluation 58 (2018) 80–96
determinant of test scores, albeit less than schools. Owens (2010) also
concludes that neighborhoods have a significant influence on educa-
tional attainment even when school characteristics are taken into
consideration. In addition, the disadvantages of students coming from
neighborhoods with a low level of socioeconomic development are not
eliminated by changing the school they attend.
In the case of Latin American, Vivas Pacheco, Correa Fonnegra, and
Domínguez Moreno (2011) study the relationship between educational
achievement and the socioeconomic and family context of Colombian
students. They conclude that there is a tight nexus between the quality
of the local context and the educational outcomes they may achieve. In
fact, of the different factors considered, the neighborhood has the
biggest impact on educational achievement. In other words, students in
disadvantaged neighborhoods are worse off than others living in better
areas. Del Valle and Fernández (2014) study this relationship, in the
case of Costa Rica, using the socioeconomic level of the community.
They conclude that there are correlations between a district’s level of
development and student outcomes. Giménez and Castro (2017) em-
phasize the importance of reaching a minimum level of public infra-
structure to guarantee that basic household needs like access to elec-
tricity and telecommunications are fully covered. These authors argue
that the quality of these types of services are strongly related to student
achievement in Costa Rica. Jiménez, Matus, and Martínez (2014) find
that information and communication technologies (ICTs) have a posi-
tive effect on economic growth, innovation and higher-quality educa-
tion in Mexico. In addition, the importance of ICTs as a potential con-
tributor to development is explicitly recognized in the Sustainable
Development Goals.
2.3. Educational development
Disadvantaged schools are more likely to suffer from shortages or
inadequacy of educational materials and physical infrastructure—such
as school grounds and buildings, heating/cooling and lighting systems
and instructional space, such as classrooms—than more advantaged
schools (OECD, 2013). This is a critical fact, given that the schools’
capacity to satisfy students’ needs may be hampered by this factor.
For the Latin American case, Murillo and Román (2011) find that
find that the effect on the achievement of primary education may be
affected by the schools ‘infrastructure, services, and didactic facilities
(sport installations, labs, libraries and computers), although the effect
may differ from country to another. Empirical studies also show that
students tend to perform better in schools that provide a conducive
learning environment, with less likelihood of grade repetition, student
truancy or better behavior (Jennings & Greenberg, 2009).
Furthermore, many authors highlight the difficulty that schools in
disadvantaged areas have to attract and retain good teachers—a key
component of the education process (Bruns & Luque, 2014; Pelayo &
Brewer, 2010). Lupton (2004) argues that schools in disadvantaged
areas of the UK have greater difficulties to recruit experienced teachers.
In addition, these teachers face a much more challenging environment,
lowering teacher quality and, in consequence, educational quality. In
developing countries, UNESCO (2014) warns of unequal teacher dis-
tribution in schools depending on the socioeconomic characteristics and
its consequences on rising educational inequality. Evans and Popova
(2015) recommend pedagogical interventions (including computer-as-
sisted learning) that should be implemented according to the students’
the specific needs.
2.4. Electoral participation and social capital
Academic literature has not focused on the influence of electoral
participation on student performance. However, we consider this could
be important because a high level of electoral abstentionism might re-
flect significant processes of social self-exclusion. Research on political
behavior has examined, mostly in the United States, the effect of
G. Gimenez et al.
education on community and electoral participation. Dee (2010),
Shields and Goidel (1997), Verba, Schlozman, and Brady (1995) or
Wolfinger and Rosenstone (1980), for instance, conclude that education
is a strong predictor of political participation. Hillygus (2005, p. 28)
claims that “civic education expands the capacity of citizens to engage
in self-rule by teaching citizens the behaviors and knowledge necessary
to identify political preferences, and pursue political interests”. Social
sciences and humanities college students tend to participate more in
political activities (Nie & Hillygus, 2001).
With respect to social capital, Grootaert (2001) and Grootaert and
van Bastelaer (2001) consider it to be a potential contributor to in-
creasing welfare and capabilities in poor countries. Stronger social
networks reduce social pathologies and negative emotions that hinder
students’ achievement (Muennig, 2010). Masino and Niño-Zarazúa
(2016) consider “participatory and community management interven-
tions” as underlying drivers of change that facilitate improving edu-
cation quality in developing countries. It may be hypothesized that
parents’ community engagement is positively related to their degree of
involvement in their children’s school and learning, which, in turn, will
have an impact (presumably, positive) on their performance.
From a sociological perspective, Bourdieu (1980) pioneered the
notion of social capital in the field, referring to the set of resources that
one can mobilize through one’s friends, relatives or acquaintances.
Other authors, such as Coleman (1988) and Schuller (2001), elaborated
on the relationship between social capital and human capital,.More
recently, there has been increasing interest in studying this relationship,
both theoretically and empirically (Acar, 2011; Behtoui & Neergaard,
2015).
2.5. Multidimensional perspective
Over the years, the focus of the literature on development has
shifted away from a one-dimensional measure, based on income, to a
multidimensional perspective, called human or social development,
which has become the central concept of the global development
agenda in recent decades. This is evident in the Millennium Development
Goals or, more recently, in the new Sustainable Development Goals fra-
mework (Addison, Niño-Zarazúa, & Tarp, 2015).
In this paper, we consider social development to be a concept that
goes far beyond economic aspects, emphasizing the realization of basic
needs and human potential (Nahar, 2015; Parker, 2005). In line with
Sen (1999) and the United Nations Development Programme approach,
the focus of social development is on enhancing capabilities and
widening human choices. In other words, it is about empowering
people, particularly the most vulnerable in society (World Bank, 2005).
The empirical work presented below aims to contribute to the de-
bate on the determinants of student achievement in developing coun-
tries by introducing the dimension of social development. It in-
corporates key dimensions for the quality of life of individuals and
groups as well as economic factors, such as health, education and po-
litical participation. Furthermore, it focuses on the level of social de-
velopment of the school district, in contrast to the usual focus on the
district where the children reside. Sampson et al. (2002) highlight that
teenagers interact with each other in many different contexts outside of
home. Therefore, the common practice in neighborhood-effects re-
search of looking solely at the characteristics of the individual's place of
residence suffers from the problem that many activities of interest un-
fold in places —e.g. schools, parks, center city areas— outside of the
residential neighborhoods in which the individuals involved in these
activities live. As students spend a significant part of their time in these
places, the districts where schools are located become areas of parti-
cular interest. In addition to school hours, numerous extracurricular
and leisure activities are carried out in the area around the school.
This paper tries to verify if student performance is greater in the
schools situated in districts where the SDI is higher. That is, in districts
where the population enjoy a healthy life, where citizens achieve a
82
Studies in Educational Evaluation 58 (2018) 80–96
decent standard of living, where students have access to education
services that favor the development of their human capital and where
the inhabitants have a feeling of responsible active participation.
3. Data and empirical model
3.1. Data
Our data set comprises two different sources: the PISA 2012 data-
base and the SDI elaborated by the Costa Rican Ministerio de
Planificación Nacional y Política Económica (MIDEPLAN). The geoloca-
tion of the participating schools, supplied by the Costa Rican Ministerio
de Educación and the Estado de la Educación, allowed us to link the PISA
dataset with the SDI for the 473 districts, the smallest administrative
unit into which the country is divided.2 Combining PISA 2012 with
local microdata like the SDI is a novelty.
The PISA 2012 database provides us with standardized student test
scores in math, reading and science, which are used as the proxy of
educational performance. In addition, it offers wide-ranging data at
student, family and school level, which will be used as control vari-
ables.
When quantifying the social development of the school’s district, we
use the SDI. According to the MIDEPLAN report (2013: pp. 15–16), the
index is built on the possibilities of the population to access and enjoy a
set of basic rights, classified into four dimensions:
• Health: possibility of enjoying a healthy life, which implies having
access to formal health care, guaranteeing an adequate quality of
life.
• Economy: a decent standard of living through earning income from
participating in economic activity.
• Education: adequate access to educational and training services that
contribute to the development of human capital.
• Electoral participation: development of the sense of belonging and
social cohesion among the population, reflected through participa-
tion in national and local civic procedures.
The SDI is built through a set of 11 socioeconomic indicators
grouped in these 4 dimensions (see Fig. 1). As the indicators are in
different units and scales, they first need to be normalized according to
two formulas.
If a higher value of the indicator shows a better situation in terms of
development, the formula applied is:
[x i − x min ]
yi = 100
[x max − x min ] (1)
If a lower value indicates a worse situation, then the formula ap-
plied is:
[x max − x i ]
yi = 100
[x max − x min ] (2)
where Yi is the indicator, xi its mean, and xmax and xmin its maximum
and minimum values.
The SDI is constructed as the weighted sum of the indicators,
grouped to the four categories. The weights are assigned by a principal
component analysis —a data reduction technique that takes into
2
In Costa Rica, private and public education (92. 1% of the total schools in Costa Rica
in 2011) coexist and are regulated and supervised by the Ministerio de Educación. The
educational system is structured in four levels: pre-school (before 7 years old), primary
education (from 7 to 12), secondary school (from 12 to 18) and higher education.
Secondary education is divided into two cycles of three years. The first cycle is dedicated
to general education; the second cycle keeps a core curriculum but implies a specializa-
tion —academic or technical. During this last cycle, known as diversified education,
students can choose various instructional modalities with different durations and peda-
gogical approaches.
G. Gimenez et al. Studies in Educational Evaluation 58 (2018) 80–96

Fig. 1. Social Development Index.

Notes: The information used in the ‘education’ dimension refers to primary education in public schools, which represented 92.1% of the total schools in Costa Rica in
2011.
Source: Prepared by the authors with information from MIDEPLAN (2013).

Table 1 information of the original set. The SDI shows values between 0 and
Missing data analysis. 100. Values close to 100 indicate a better situation.
Variable Missing Total Share of missing SDI = I j ∙αj (3)
data
where I and α refer to the indicator and weight, respectively, and the
Share of government funding 492 4,602 10.69% subscript j corresponds to each of the four dimensions in Fig. 1.
Share of fully certified teachers at school 1,495 4,602 32.49%
Finally, a key concern with the data is the existence of endogeneity
Shortage of math teachers 36 4,602 0.78%
Shortage of reading teachers 62 4,602 1.35% problems that can give rise to biased estimations. These problems may
Shortage of science teachers 36 4,602 0.78% have their origin in: a) measurement errors; b) the omission of some
Non-native student 43 4,602 0.93% relevant variables, correlated with observed variables, which are used
Books at home 197 4,602 4.28%
as explanatory; and c) the use of explanatory variables that are not
Parents' education 331 4,602 7.19%
Parents' occupation 416 4,602 9.04%
strictly exogenous due to the presence of simultaneity.
The presence of endogeneity would imply that the explanatory
Total missing 3108
Total data used in baseline model 69,030
variables and the error term are correlated. If this were the case, the use
regression of instrumental variables would provide unbiased and consistent esti-
mators, which would be more efficient. To verify the presence of en-
Share of missing values imputed 4.50%
dogeneity in the model, we focus on the variable SDI, which is the key
element in our empirical study. A Wooldridge (1995) robust score test
account the structure and variation of the data to assign the weights. for endogeneity, with the null hypothesis of the SDI being exogenous, is
Through this technique, the information contained in the set of in- performed. The test is carried out making use of a model with instru-
dicators can be reduced to obtain a single index that carries most of the mental variables (2SLS estimation with a robust VCE) in which
household income at the district level is used as the instrument for SDI.

83
G. Gimenez et al. Studies in Educational Evaluation 58 (2018) 80–96

Table 2
Estimates of education production function with Social Development Index. Dependent variable: PISA math scores.
(1) (2) (3) (4) (5) (6)
Control SDI Health Economy Educatio- Electoral participation
n

Male 24.128*** 24.126*** 24.127*** 24.126*** 24.123*** 24.126***

(1.765) (1.765) (1.765) (1.765) (1.765) (1.765)

Age 11.796*** 11.799*** 11.795*** 11.797*** 11.798*** 11.798***

(3.354) (3.353) (3.354) (3.353) (3.353) (3.353)

Non-native student −8.654 −8.674 −8.652 −8.680 −8.664 −8.641

(5.292) (5.294) (5.292) (5.293) (5.292) (5.293)

Books at home
11–25 books 10.046*** 10.040*** 10.047*** 10.039*** 10.042*** 10.052***
(2.392) (2.392) (2.392) (2.392) (2.392) (2.392)
26–100 books 14.160*** 14.151*** 14.159*** 14.148*** 14.151** 14.161***
(2.726) (2.726) (2.726) (2.726) (2.725) (2.726)
101–200 books 18.862*** 18.850*** 18.863*** 18.847*** 18.853*** 18.867***
(4.371) (4.371) (4.371) (4.371) (4.371) (4.371)
201–500 books 22.010** 21.997** 22.011** 21.990** 21.997** 22.029**
(9.079) (9.082) (9.079) (9.082) (9.080) (9.082)
More than 500 books 15.351 15.346 15.350 15.337 15.348 15.363
(11.222) (11.221) (11.221) (11.221) (11.221) (11.224)

Parents' occupation 0.197*** 0.196*** 0.197*** 0.196*** 0.196*** 0.197***

(0.045) (0.045) (0.045) (0.045) (0.045) (0.045)

Parents' education
6 9.739** 9.724** 9.740** 9.734** 9.721** 9.733**
(4.494) (4.493) (4.493) (4.493) (4.492) (4.493)
11 15.592* 15.554* 15.592* 15.563* 15.557* 15.579*
(8.032) (8.032) (8.032) (8.031) (8.031) (8.031)
12 17.882*** 17.841*** 17.886*** 17.856*** 17.839*** 17.879***
(4.389) (4.389) (4.389) (4.389) (4.389) (4.388)
14 2.670 2.653 2.672 2.660 2.654 2.666
(5.010) (5.009) (5.010) (5.009) (5.008) (5.009)
16 18.103*** 18.066*** 18.107*** 18.081*** 18.068*** 18.099***
(4.124) (4.125) (4.125) (4.125) (4.125) (4.124)

Number of students 0.005 0.004 0.005 0.004 0.004 0.005

(0.005) (0.005) (0.005) (0.005) (0.005) (0.005)

Privately operated 14.922 11.092 15.327 11.205 12.709 16.015

(21.156) (21.176) (21.006) (21.291) (20.317) (21.194)

Share of government funding −0.008 −0.008 −0.008 −0.008 −0.008 −0.009

(0.079) (0.079) (0.079) (0.079) (0.079) (0.079)

School's community location

Small Town 6.091 2.078 6.576 3.000 1.969 7.511
(6.472) (6.223) (6.545) (6.420) (6.228) (6.669)
Town 12.088* 3.797 12.940* 5.497 4.505 13.151**
(6.748) (6.703) (6.852) (6.842) (6.568) (6.709)
City 6,487 1,244 7,176 1,669 2,277 8,998
(9142) (8644) (9352) (8713) (8519) (9540)
Large City 28,817 19,601 29,487 19,127 21,858 32,247
(25,212) (26,036) (24,641) (26,215) (24,866) (24,671)
Share of fully certified teachers at school −8.451 −8.613 −8.442 −8.549 −8.643 −8.488
(6.515) (6.506) (6.516) (6.505) (6.504) (6.510)

Shortage of math teachers

Very little −6.618 −6.545 −6.817 −6.787 −6.678 −7.142
(8.959) (8.758) (9.073) (8.829) (8.702) (8.854)
To some extent −0.126 −1.155 −0.135 −0.988 −1.223 −0.218
(12.694) (12.269) (12.647) (12.386) (12.167) (12.470)
A lot −18.520 −10.716 −19.356 −13.909 −8.093 −21.212*
(12.913) (12.528) (13.211) (13.603) (11.649) (12.889)

School autonomy 16.585** 15.755* 16.544** 16.064** 15.249* 16.217*

(8.374) (8.206) (8.314) (8.124) (8.018) (8.435)

SDI 0.616***
(0.193)
Health −0.108
(0.222)
Economy 0.381**
(0.168)
Education 0.610***
(0.150)
(continued on next page)

84
G. Gimenez et al.
Table 2 (continued)
(1) (2)
Control SDI
Electoral participation
Constant 183.969*- 148.840*-
** **
(54.591) (55.221)
Notes: The estimates with plausible values are made using the PV Stata module developed by Macdonald (2014). Regressions weighted by students' sampling
probability. Robust standard errors adjusted for clustering at district level are in parentheses.
*** ρ < 0.01.
** ρ < 0.05.
* ρ < 0.1.
2
The result of the test, with χ(1) = 2.49, leads us to accept the null hy-
pothesis: the SDI is exogenous in the model. In addition, a weak iden-
tification test value of 33.37 lets us conclude that the instrument used is
strong, below the critical values for 2SLS relative biases tabulated by
Stock and Yogo (2005). When we carry out the endogeneity analysis
with the four dimensions of the SDI we obtain the same conclusions,
with higher levels of acceptance of the null hypothesis of them being
exogenous.
3.2. Empirical model
The proposed model, which tests the hypothesis that social devel-
opment significantly influences educational performance, is formulated
as follows:
Yijk = β0 + β1 X1ijk + β2 X2jk + β3 SDI3k + μ3k SDI3k + εijk
β0 = γ00 + v0k + u 0jk
Eq. (4) denotes the basic education production function. These func-
tions facilitate the adoption of decisions about what resources should be
used, and in what amount, in order to optimize the learning process for
students. To do so, they empirically establish, through regression
techniques, the relationship between the student’s outcomes (output),
often test scores, and a set of student, family and school characteristics
(inputs).
In our function, Yijk is the expected test score of student i enrolled in
school j in district k. X1ijk and X2jk are vectors of control variables at the
individual and school level, respectively. SDI3k stands for the Social
Development Index, and εijk is the unexplained component for student i
in school j in district k.
Under the education production function framework, a set of var-
ious educational inputs have been identified in the empirical literature
(Harris, 2010). These are usually grouped into three categories: 1) in-
dividual characteristics of the student; 2) characteristics of the student’s
household; and 3) characteristics of the school. This paper aims to go
beyond these three categories by considering the characteristics of the
school district. The control variables are introduced following the
model of Hanushek, Link, and Woessmann (2013) as the benchmark.
Among the individual characteristics, gender, age and country of birth
are selected. PISA results show that, in most countries, girls outperform
boys in reading, but underperform boys in math (OECD, 2015). The
students’ age (PISA students are aged between 15 years and 3 months
and 16 years and 2 months at the time of the assessment) is also con-
sidered as a factor to control for (OECD, 2014). Finally, students born
abroad tend to obtain lower scores than national students, although
these differences can vary across school systems (OECD, 2012).
The family variables are parental education (highest educational
level attained), their occupation and the number of books at home (as a
Studies in Educational Evaluation 58 (2018) 80–96
(3) (4) (5) (6)
Health Economy Educatio- Electoral participation
n
0.598**
(0.288)
191.541*- 174.082*- 143.154*- 156.608***
** ** **
(57.067) (54.380) (54.966) (56.339)
proxy of the socioeconomic and cultural level). Socioeconomic char-
acteristics are strong influencing factors on academic achievement,
both in developed and developing countries. A substantial number of
studies have shown that parents’ education, parents’ occupation, and
household resources are positively associated with performance in
European countries (Ammermuller, Heijke, & Wößmann, 2005; Schutz,
Ursprung, & Wößmann, 2008; Wößmann, 2005); the United States
(Hanushek, 2002; Tian, 2006); Asia (Wößmann, 2003a) and Latin
America (Ayala, Marrugo, & Saray, 2011; Castro Aristizábal, Gimenez,
& Pérez Ximénez-de-Embún, 2018; Giménez & Castro, 2017;
Hernandez-Zavala, Patrinos, Sakellariou, & Shapiro, 2006).
The school characteristics are school size (number of students en-
rolled at school), school ownership, the proportion of public funding,
school location, the proportion of certified teachers, the shortage of
teachers in each area, and school autonomy. School characteristics
(4) play, as a whole, an important role in educational performance
(Hanushek & Wößmann, 2007; Lee & Barro, 2001; Wößmann, 2003b).
(5) In one third of the countries participating in PISA, the effect of school
characteristics is greater than that of student-level variables (OECD,
2003). A substantial proportion of the variation of test scores within
countries participating in PISA is associated with the school students
attend and with what the schools do (Freeman and Viarengo, 2014).
These results have also been found specifically in the Latin American
countries participating in PISA (Castro Aristizábal, Gimenez, & Pérez
Ximénez-de-Embún, 2017, 2018).
Finally, as the effect of SDI can differ in each district, μ3k considers
this heterogeneity, allowing the SDI to have different slopes and in-
tercepts for each district. This coefficient is assumed to be normally
distributed with mean 0 and uncorrelated with εijk . See Tables A1 and
A2 for a detailed description of these variables and sources, as well as
the statistics.
In Eq. (5), u 0jk and v0k are the respective deviation of schools’ and
districts’ mean from the overall mean (γ00 ). They are assumed to be
normally distributed with mean 0, and uncorrelated with εijk . Control-
ling for school and district effects should mitigate the bias from dif-
ferences that are correlated with test scores.
The data present a nested structure in which students are grouped
into schools, and schools are grouped into districts. Originally, districts
are not sampled in PISA data in Costa Rica, and these do not include
information about the location of the schools in the districts. Therefore,
we incorporate this information into the original data. The hierarchical
structure implies that the students’ data are not randomly distributed;
students of the same school are expected to share more similarities
among each other than with students from other schools. Consequently,
techniques like OLS are not appropriate to estimate the model as the
presence of unobserved heterogeneity is a fact.
For this reason, we use hierarchical models which allow us to deal
with unobserved differences by modeling district- and school-specific
85
G. Gimenez et al. Studies in Educational Evaluation 58 (2018) 80–96

Table 3
Estimates of education production function with Social Development Index. Dependent variable: PISA reading scores.
(1) (2) (3) (4) (5) (6)
Control SDI Health Economy Education Electoral participation

Male −24.159*** −24.165*** −24.159*** −24.162*** −24.167*** −24.160***

(1.973) (1.973) (1.973) (1.973) (1.973) (1.973)

Age 5.714 5.720 5.714 5.717 5.721 5.715

(4.217) (4.216) (4.218) (4.217) (4.216) (4.217)

Non-native student −10.968 −10.986 −10.969 −10.998 −10.971 −10.959

(6.693) (6.692) (6.693) (6.693) (6.690) (6.696)

Books at home
11–25 books 6.142** 6.132** 6.141** 6.131** 6.136** 6.147**
(2.758) (2.758) (2.758) (2.758) (2.758) (2.759)
26–100 books 6.292** 6.280** 6.292** 6.278** 6.283** 6.294**
(2.955) (2.955) (2.955) (2.954) (2.955) (2.955)
101–200 books 14.244*** 14.226*** 14.243*** 14.224*** 14.236*** 14.248***
(4.821) (4.822) (4.821) (4.822) (4.821) (4.821)
201–500 books 15.585* 15.560* 15.583* 15.557* 15.569* 15.595*
(8.223) (8.228) (8.223) (8.226) (8.226) (8.226)
More than 500 books 15.102 15.120 15.097 15.108 15.132 15.124
(10.297) (10.292) (10.296) (10.296) (10.288) (10.297)

Parents' occupation 0.191*** 0.190*** 0.191*** 0.190*** 0.190*** 0.191***

(0.053) (0.053) (0.053) (0.053) (0.053) (0.053)

Parents' education
6 14.566** 14.549** 14.565** 14.561** 14.544** 14.561**
(5.864) (5.864) (5.864) (5.863) (5.862) (5.865)
11 8.566 8.471 8.554 8.527 8.493 8.542
(9.546) (9.549) (9.547) (9.546) (9.543) (9.547)
12 27.393*** 27.337*** 27.389*** 27.360*** 27.334*** 27.390***
(5.363) (5.361) (5.362) (5.362) (5.360) (5.363)
14 9.500* 9.470* 9.497* 9.483* 9.472* 9.497*
(5.221) (5.221) (5.220) (5.221) (5.220) (5.221)
16 24.744*** 24.694*** 24.741*** 24.713*** 24.695*** 24.738***
(5.401) (5.400) (5.401) (5.401) (5.400) (5.402)

Number of students 0.002 0.001 0.002 0.001 0.001 0.002

(0.008) (0.008) (0.008) (0.008) (0.008) (0.008)

Privately operated 30.294 24.065 29.690 25.016 26.044 31.113

(20.026) (19.912) (20.388) (19.774) (19.005) (20.218)

Share of government funding 0.007 0.007 0.007 0.006 0.007 0.006

(0.110) (0.110) (0.110) (0.110) (0.110) (0.110)

School's community location

Small Town 15.474** 9.557 14.819* 11.369 9.607 16.469**
(7.656) (7.271) (7.682) (7.395) (7.221) (7.865)
Town 18.214** 5.479 17.078** 9.028 6.754 18.913**
(8.097) (7.610) (7.846) (7.736) (7.745) (8.154)
City 9.861 1.894 8.992 3.118 3.571 11.569
(10.892) (10.725) (11.096) (10.759) (10.310) (11.182)
Large City 17.744 3.568 16.780 4.322 7.248 20.120
(26.556) (27.408) (26.829) (28.010) (25.634) (26.232)
Share of fully certified teachers at school −12.500* −12.720* −12.505* −12.635* −12.761* −12.526*
(7.142) (7.132) (7.142) (7.135) (7.132) (7.141)

Shortage of reading teachers

Very little 7.428 6.525 7.583 6.588 6.598 6.861
(7.521) (7.303) (7.494) (7.424) (7.234) (7.585)
To some extent 4.151 4.476 4.473 3.885 4.113 3.926
(20.456) (19.984) (20.605) (20.189) (19.777) (20.360)
A lot 14.867 15.564 14.956 14.700 15.291 15.698
(19.758) (19.221) (19.735) (19.384) (19.032) (19.935)

School autonomy 12.717 11.511 12.832 11.937 10.924 12.354

(8.217) (8.007) (8.285) (7.916) (7.763) (8.279)
SDI 0.890***
(0.208)
Health 0.140
(0.240)
Economy 0.516***
(0.168)
Education 0.871***
(0.161)
Electoral participation 0.409
(0.312)

(continued on next page)

86
G. Gimenez et al.
Table 3 (continued)
(1) (2)
Control SDI
Constant 323.903*** 273.759***
(70.709) (69.521)
Notes: The estimates with plausible values are made using the PV Stata module developed by Macdonald (2014). Regressions weighted by students' sampling
probability. Robust standard errors adjusted for clustering at district level are in parentheses.
*** ρ < 0.01.
** ρ < 0.05.
* ρ < 0.1.
factors (Raudenbush & Bryk, 2002). With this method, unbiased esti-
mators are obtained because the potential correlations between the
error terms among students of the same school and district are taken
into account.
Finally, we have to point out that the presence of missing ob-
servations represents a major problem when working with PISA data.
Table 1 shows the missing value analysis of our dataset. 4.5% of the
observations are missing, which apparently is a low percentage.
Nevertheless, the deletion of all the students that have a missing value
for at least one variable would reduce the sample size from 4,602 to
2,381 observations. This can generate biases that affect statistical in-
ference.
To circumvent this issue, we use the Bayesian bootstrap method of
Rubin and Schenker (1986), implemented by Mander and Clayton
(2007). Each missing value is substituted with the value from the most
similar case for which the variable is not missing. The approach has
some advantages over other imputation methods that make it appro-
priate in this case. Durrant (2009) and Medina and Galván (2007)
highlight these advantages: 1) the method is robust to model mis-
specifications; 2) it avoids distributional assumptions; 3) it uses oc-
curring values for imputation; and 4) it is suitable for dealing with
categorical data.
4. Results of the estimation of the education production function
The empirical model is estimated for math, reading and science
(Tables 2–4). In each table, Column 1 presents the estimations without
the SDI; Column 2 shows the results with the SDI; and Columns 3–6
contain the estimations with the four dimensions of the mentioned
index: health, economy, education and electoral participation.
The estimations in Column 1 in the three tables suggest that—in the
case of student and family factors—gender, age (except for reading),
books at home, parental occupational status and parental education are
correlated with test scores. Regarding school characteristics, school
autonomy is positive and significant in math and science and share of
fully certified teachers is negative and significant in reading and sci-
ence. The share of fully certified teachers in Costa Rica is 80%. This
figure is close to the average of OECD countries (87%) and much higher
than that of other Latin American countries participating in PISA.
Differences between schools are small and this might be behind the
inverse correlation with the students’ scores. Individual and family
variables are more clearly linked with student achievement than school
variables, which is consistent with the PISA literature. In fact, the lack
of significance of most school resources and characteristics is common
to other studies, but it must not be interpreted as meaning that
schooling does not affect cognitive abilities (OECD, 2014).
Results support the hypothesis on the effect of social environment
on academic achievement. Column 2 of Tables 2–4 shows that the SDI
has a significant positive effect on the test scores of the three subjects.
The higher the social development of the school district, the better the
students’ achievement.
The disaggregation of the SDI (Columns 3–6) shows mixed results.
While the dimensions of economy and education have a significant and
87
Studies in Educational Evaluation 58 (2018) 80–96
(3) (4) (5) (6)
Health Economy Education Electoral participation
314.104*** 310.722*** 266.209*** 305.162***
(69.866) (70.279) (70.317) (70.352)
positive correlation with the test scores in all subjects, the same cannot
be said for health and electoral participation (these last two dimensions
have the lowest correlation with the tests scores and have a lower
weight in the construction of the SDI than the other two dimensions).
The health subindex, which combines four different indicators:
births to unmarried teenage mothers, low weight problems in children,
infant mortality and quality of potable water, has no statistical sig-
nificance in any of the regressions. This result is consistent with recent
literature. Aguilera and Quintana (2011), for instance, did not find a
clear and robust relationship between body weight and academic
achievement in their study on Mexico. Furthermore, in an analysis of
systematic reviews on the effectiveness of different types of programs to
improve learning at the primary level in developing countries, Evans
and Popova (2015) find a questionable effect of health programs. In
Conn (2014) and McEwan (2015), these programs are the least effective
in terms of improving test scores.
Ozier (2014), however, gives a more positive view of the effect of
health on educational outcomes. He finds significant positive ex-
ternalities of community health programs in early childhood in Kenya.
As Conn (2014) shows, these interventions may contribute to better
direct cognitive outcomes, such as attention and memory. Miguel and
Kremer (2004) also documented the positive effect of health programs
on increasing school enrollment and attendance, but not on better
academic outcomes.
Thus, our results should not be interpreted as meaning that im-
proving the health environment of the district is irrelevant for educa-
tional performance. The literature has not been able to find robust
evidence of the effect of health interventions on academic test scores.
One explanation may be that improving health—thereby likely im-
pacting school attendance and cognitive outcomes positively— is a
necessary, but not a sufficient condition, for better academic outcomes.
If the quality of instruction is as poor as before the intervention, it will
not make a big difference. Another explanation could have to do with
the nature of the data. Firstly, in contrast to some of its regional peers,
Costa Rica is a middle-income country where basic health problems, as
well as health inequalities, are not widespread. According to the Human
Development Index data (United Nations Development Programme,
2015), Costa Rica compares remarkably well in terms of health in-
dicators with the Latin American region, as well as with countries of
similar human development.3 Secondly, we do not control for the dis-
trict in which a child lives (this information is not available), only for
the one in which the school is situated. This could contribute to ex-
plaining the statistical insignificance of the health index because health
care conditions (such as drinking water) may be more relevant where a
child lives than where a child attends school.
The economic dimension is composed of two variables, namely,
3
For instance, life expectancy at birth is 79.6 years in Costa Rica versus 75 years in the
Latin America & the Caribbean region, and 75.1 years in the group of high human de-
velopment to which it belongs. The percentage of malnourished children in the country is
only 5.6%, less than half the regional average (13.9%). Infant mortality rates are also
substantially lower (8.4 vs. 15.2 per 1,000 live births in the region). Furthermore, Costa
Rica shows less inequality in life expectancy: 7.3% versus 13.3% in the region.
G. Gimenez et al. Studies in Educational Evaluation 58 (2018) 80–96

Table 4
Estimates of education production function with Social Development Index. Dependent variable: PISA science scores.
(1) (2) (3) (4) (5) (6)
Control SDI Health Economy Education Electoral participation

Male 12.299*** 12.296*** 12.300*** 12.297*** 12.292*** 12.297***

(2.775) (2.775) (2.775) (2.775) (2.775) (2.775)

Age 8.590** 8.597** 8.590** 8.593** 8.600** 8.593**

(3.571) (3.570) (3.571) (3.571) (3.570) (3.571)

Non-native student −7.072 −7.085 −7.071 −7.084 −7.089 −7.067

(6.680) (6.681) (6.679) (6.681) (6.678) (6.681)

Books at home
11–25 books 11.454*** 11.446*** 11.453*** 11.448*** 11.447*** 11.458***
(3.860) (3.860) (3.860) (3.860) (3.860) (3.860)
26–100 books 16.865*** 16.856*** 16.866*** 16.857*** 16.856*** 16.867***
(3.604) (3.603) (3.604) (3.603) (3.603) (3.604)
101–200 books 16.318*** 16.303*** 16.317*** 16.308*** 16.309*** 16.325***
(4.611) (4.611) (4.611) (4.612) (4.610) (4.611)
201–500 books 24.048** 24.031** 24.048** 24.029** 24.032** 24.058**
(9.920) (9.922) (9.921) (9.923) (9.919) (9.922)
More than 500 books 16.814* 16.813* 16.814* 16.806* 16.824* 16.826*
(9.264) (9.261) (9.264) (9.263) (9.258) (9.267)

Parents' occupation 0.169*** 0.168*** 0.169*** 0.169*** 0.168*** 0.169***

(0.048) (0.048) (0.048) (0.048) (0.048) (0.048)

Parents' education
6 8.283 8.268 8.282 8.280 8.262 8.277
(5.784) (5.785) (5.785) (5.784) (5.783) (5.784)
11 11.249 11.181 11.246 11.217 11.171 11.238
(9.672) (9.671) (9.673) (9.672) (9.668) (9.672)
12 16.857*** 16.814*** 16.855*** 16.836*** 16.806*** 16.855***
(6.098) (6.099) (6.099) (6.099) (6.098) (6.097)
14 3.047 3.031 3.045 3.039 3.033 3.047
(6.440) (6.440) (6.440) (6.440) (6.439) (6.439)
16 15.922** 15.882** 15.920** 15.902** 15.880** 15.917**
(6.676) (6.679) (6.678) (6.677) (6.677) (6.676)

Number of students 0.006 0.005 0.006 0.005 0.005 0.006

(0.005) (0.005) (0.005) (0.005) (0.005) (0.005)

Privately operated 29.519 24.762 29.176 26.216 25.879 30.776

(18.828) (18.736) (18.916) (18.749) (17.923) (19.085)
Share of government funding 0.033 0.033 0.033 0.033 0.033 0.032
(0.102) (0.102) (0.102) (0.102) (0.102) (0.102)

School's community location

Small Town 5.230 0.996 4.868 2.861 0.753 6.520
(7.683) (7.255) (7.611) (7.410) (7.183) (7.935)
Town 10.640 1.178 10.026 5.110 1.467 11.549
(7.041) (7.048) (6.920) (7.076) (6.975) (7.098)
City 4.449 −1.018 3.990 0.540 0.056 6.760
(10.263) (9.796) (10.349) (9.995) (9.495) (10.421)
Large City −0.771 −11.235 −1.250 −8.842 −9.244 2.258
(26.946) (27.304) (26.984) (27.596) (26.123) (26.368)
Share of fully certified teachers at school −11.171* −11.362* −11.176* −11.260* −11.428* −11.195*
(6.737) (6.726) (6.738) (6.731) (6.719) (6.733)

Shortage of science teachers

Very little 11.293 10.568 11.414 10.771 10.133 10.681
(10.211) (10.034) (10.202) (10.174) (9.956) (10.226)
To some extent −1.246 −0.152 −1.175 −0.909 0.285 −1.054
(16.057) (15.896) (16.053) (16.009) (15.781) (15.980)
A lot 37.085 32.521 36.795 33.819 34.534 37.041
(26.764) (28.012) (26.841) (28.073) (26.413) (26.752)

School autonomy 14.493* 13.507* 14.548* 14.055* 12.897* 13.983*

(7.809) (7.644) (7.847) (7.615) (7.449) (7.915)
SDI 0.712***
(0.223)
Health 0.081
(0.217)
Economy 0.327*
(0.173)
Education 0.763***
(0.169)
Electoral participation 0.525
(0.320)

(continued on next page)

88
G. Gimenez et al. Studies in Educational Evaluation 58 (2018) 80–96

Table 4 (continued)

(1) (2) (3) (4) (5) (6)

Control SDI Health Economy Education Electoral participation

Constant 255.660*** 214.990*** 249.954*** 247.140*** 204.574*** 231.365***

(57.862) (58.224) (57.844) (57.933) (57.550) (60.730)

Notes: The estimates with plausible values are made using the PV Stata module developed by Macdonald (2014). Regressions weighted by students' sampling
probability. Robust standard errors adjusted for clustering at district level are in parentheses.
*** ρ < 0.01.
** ρ < 0.05.
* ρ < 0.1.

Table 5 electricity consumption and access to Internet correlate positively with

Impact on test scores points of 1 standard deviation change. educational achievement, as the results suggest.
Math Reading Science
The education dimension includes four school inputs which are
different from the ones in the estimated education production function.
Aggregate SDI 7.975 11.521 9.223 They are educational infrastructure, special education programs, tea-
Health n.s. n.s. n.s. chers who give classes to heterogeneous students, and school failure. As
Economy 6.492 8.790 5.575
Education 8.649 12.360 10.825
expected, districts that score higher in this dimension tend to have
Electoral participation 5.394 n.s. n.s. students who perform better in test scores. This is consistent with the
meta-review in Evans and Popova (2015), who find pedagogical in-
Note: n.s. means non-significant at the 10 percent level. terventions to be highly effective, as long as they are tailored to the
specific needs of the students and are well implemented.
Table 6 Finally, in the case of electoral participation, the estimations show a
Estimates of education production function with Social Development Index significant effect only when it is measured against math test scores, but
including a quadratic term. Dependent variable: PISA math scores. not against reading and science test scores. So, the conventional view of
(1) (2) (3) (4) (5) the positive relationship between economic development and civic
SDI Health Economy Education Electoral participation summarized above does not hold in Latin America. In fact,
participation the reverse seems to be true: hostile economic environments foster
SDI 2.758** participation in civic organizations, which is used as a means to over-
(1.368) come the lack of societal resources (Gundelach, 2016).4
SDI_sqr −0.017 Our results may be contingent on the measurement of the variable:
(0.011) MIDEPLAN (2013) considers it a proxy of social capital, but only
Health −1.345
(3.203)
measures formal participation in the political process —due to lack of
Health_sqr 0.008 data availability—. While this type of political participation could be
(0.020) the minimum threshold of community engagement, social capital is a
Economy 1.705*** multi-faceted concept that takes in many different forms and activities
(0.547)
beyond politics, such as affiliation to associations or friendship and
Economy_sqr −0.015***
(0.006) kinship relations (Bjørnskov, 2006). Furthermore, electoral turnout
Education −1.508 does not fully capture the concept of political participation, which
(1.310) should also include diverse activities aimed to influence political au-
Education_sqr 0.016 thorities led by social movements and popular organizations (Klesner
(0.010)
et al., 2007). Thus, political participation, as measured by electoral
Electoral 5.316***
participation turnout, does not necessarily imply higher levels of social capital
(1.828) (Teney & Hanquinet, 2012).
Electoral −0.053*** Table 5 shows how an increase of one standard deviation has an
participa-
effect on test scores in our explanatory variables of interest. According
tion_sqr
(0.020)
to our calculations, an increase of one standard deviation in the SDI
adds 7.98 PISA points in math, or the equivalent of a fifth of an aca-
Notes: This table only shows the coefficients, in linear as well as squared form, demic year (every 40 points in PISA is equivalent to an academic year).
of the SDI and its subcomponents. Results of the other variables are similar to The quantitative effect is substantially higher in reading (11.52 points,
those in Table 2. The estimates with plausible values are made using the PV the equivalent of almost a third of an academic year), and in science
Stata module developed by Macdonald (2014). Regressions weighted by stu- (9.22 PISA points). The education dimension in the SDI has the highest
dents' sampling probability. Robust standard errors adjusted for clustering at effect.
district level are in parentheses.
To delve deeper into the relationship between test scores and social
The * indicates statistical significance at * 10%, ** 5%, *** 1%.
development, non-linear functional forms are employed to check for the
*
ρ < 0.1.
*** ρ < 0.01. presence of decreasing returns. From the baseline model in Section 3.2,
** ρ < 0.05. a squared term for SDI or for its subcomponents is added in each re-
gression. See Tables 6–8.
residential consumption of electricity and the share of households with In reading and science scores, the quadratic term of the SDI is
access to Internet, as a proxy of household income (MIDEPLAN, 2013).
While the former constitutes one of the five biggest spending items of
4
Costa Rican households, the latter implies access to information, which Interestingly, the categories of electoral participation and economy are not correlated
brings about new economic opportunities (World Bank, 2012). Higher with each other (0.046), and the former has a low correlation with the SDI (0.188). See
Table A3 in the Appendix A.

89
G. Gimenez et al. Studies in Educational Evaluation 58 (2018) 80–96

Table 7
Estimates of education production function with Social Development Index including a quadratic term. Dependent variable: PISA reading scores.
(1) (2) (3) (4) (5)
SDI Health Economy Education Electoral participation

SDI 3.808***
(1.333)
SDI_sqr −0.023**
(0.010)
Health 2.070
(3.312)
Health_sqr −0.013
(0.021)
Economy 2.192***
(0.672)
Economy_sqr −0.019***
(0.007)
Education −0.619
(1.316)
Education_sqr 0.011
(0.010)
Electoral participation 5.797***
(1.804)
Electoral participation_sqr −0.061***
(0.020)

Notes: This table only shows the coefficients, in linear as well as squared form, of the SDI and its subcomponents. Results of the other variables are similar to those in
Table 3. The estimates with plausible values are made using the PV Stata module developed by Macdonald (2014). Regressions weighted by students' sampling
probability. Robust standard errors adjusted for clustering at district level are in parentheses.
The * indicates statistical significance at * 10%, ** 5%, *** 1%.
*ρ < 0.1.
*** ρ < 0.01.
** ρ < 0.05.

Table 8 statistically significant, whereas in math it is not. The results of PISA

Estimates of education production function with Social Development Index 2012 for Costa Rica are 407 in math, 441 in reading and 429 in science.
including a quadratic term. Dependent variable: PISA science scores. It is reasonable that the diminishing returns appear in the disciplines
(1) (2) (3) (4) (5) that have achieved the best results (reading and science). In other
SDI Health Economy Education Electoral words, it is expected that political interventions have a greater impact
participation on performance in math, the area with the greatest gap.
The categories of the index related to the economy and electoral
SDI 3.589**
(1.402) participation are significant for both the linear and the quadratic terms.
SDI_sqr −0.022** The education dimension, which had the highest effect in the previous
(0.011) estimations, loses its significance when introducing the squared term,
Health 2.666 while the health coefficients remain non-significant.
(3.360)
Health_sqr −0.017
Given that the squared coefficients are negative, this implies that
(0.022) social development gains have diminishing returns on educational
Economy 1.617** performance. Hence, an increase of the same amount in the SDI, and its
(0.682) components, has a higher effect on student achievement in the schools
Economy_sqr −0.015**
located in the disadvantaged districts. An inverted-U relationship has
(0.007)
Education −0.947 been previously discovered for educational achievement as a function
(1.136) of wealth and welfare (see Meisenberg & Woodley, 2013; Sahlgren,
Education_sqr 0.013 2015).
(0.008) Our results are also in line with Burdick-Will et al. (2011). They
Electoral 5.277**
participation
hypothesize that the key for children’s cognitive development is to
(2.360) avoid the most severely economically-distressed neighborhoods. Their
Electoral −0.054** evidence suggests that the effect of the variation in neighborhood dis-
participa- advantage may be greater for children whose starting point is a rela-
tion_sqr
tively more disadvantaged environment. In the same vein, Galster
(0.026)
(2012) points out evidence against the presence of negative effects of a
Notes: This table only shows the coefficients, in linear as well as squared form, higher neighborhood poverty rate until it reaches more than 20 percent;
of the SDI and its subcomponents. Results of the other variables are similar to after that, the externality effects rise rapidly until the neighborhood
those in Table 4. The estimates with plausible values are made using the PV reaches approximately 40 percent poverty. Subsequent increases in the
Stata module developed by Macdonald (2014). Regressions weighted by stu- poverty population appear to have no marginal external effect.
dents' sampling probability. Robust standard errors adjusted for clustering at In short, our results highlight the importance of taking into con-
district level are in parentheses. sideration district influences in terms of social development, suggesting
The * indicates statistical significance at * 10%, ** 5%, *** 1%. the presence of positive spillover effects on educational performance.5
*ρ < 0.1.
***ρ < 0.01.
** ρ < 0.05. 5
These results could also be affected by the migratory movements within Costa Rica,
which cannot be controlled for in our dataset.

90
G. Gimenez et al. Studies in Educational Evaluation 58 (2018) 80–96

Table 9
Shapley-Shorrocks Decomposition.
Variance of the test scores

Math Reading Science

Relative value 0.268 0.240 0.229

Bootstrap std. err. 0.010 0.010 0.010
Observations 4602
Bootstrap replications 100

Shapley-Shorrocks decomposition

Math Reading Science

Variable Value In percentage Value In percentage Value In percentage

Individual characteristics 0.031 11.70% 0.035 14.44% 0.006 2.76%

Familiar characteristics 0.086 31.92% 0.070 29.27% 0.083 36.37%
School characteristics 0.120 44.79% 0.104 43.15% 0.105 46.04%
Number of students 0.006 2.37% 0.006 2.35% 0.006 2.58%
Private 0.045 16.79% 0.040 16.66% 0.037 16.27%
Government funding 0.016 6.04% 0.014 5.78% 0.013 5.87%
Village 0.003 1.27% 0.007 3.07% 0.004 1.77%
Share of certified teachers 0.003 1.10% 0.001 0.27% 0.003 1.49%
Shortage of teachers in the area (to some + a lot) 0.002 0.71% 0.005 1.95% 0.007 2.99%
School autonomy 0.044 16.51% 0.031 13.08% 0.034 15.07%
Social Development Index (SDI) 0.031 11.59% 0.032 13.13% 0.034 14.83%

TOTAL 0.268 100.00% 0.240 100.00% 0.229 100.00%

Note: decomposition calculated using the Stata module proposed by Wendelspiess Chavez Juarez and Soloaga (2014).

This is consistent with some of the literature reviewed above on the
effects of local areas (neighborhoods) on educational as well as other
social outcomes. Strategies to improve the performance of students
from poorer socio-economic backgrounds will be more likely to be ef-
fective if they incorporate the goal of reducing social development by
investing in the more disadvantaged districts.
5. Estimation of the SDI contribution to the variance of test scores
As we have explained, differences in PISA scores can be due to
differences in the efforts of students or to factors that they cannot
control: individual, family or school characteristics and social devel-
opment.
Let s be the PISA test scores and C a matrix of factors defined in the
education production function (3). The expected conditional score will
be:
ŝ = E[s| C]
1 N
From (3), the variance of the test scores, (s i ∑
N i=1
−s)2 ,
can be de-
composed into its sources by estimating the relative importance of each
factor. To do this, we use the Shapley-Shorrocks decomposition. This
method is based on the calculation of the variance for all possible
permutations of the factor variables. The methodology has two im-
portant advantages over other decompositions: it is order-independent
and the components add up to the total value of the variance
(Wendelspiess Chavez Juarez & Soloaga, 2014). Its shortcoming is that
the procedure is very computation-intensive.
As the PISA test scores are continuous variables with an arbitrary
mean and dispersion, Ferreira and Gignoux (2014) argue that the best
option to estimate (6) is to use linear regressions. They also highlight
that, although the observed matrix C of factors is economically exo-
genous, it may not be exogenous in the econometric sense that its
components may be correlated with other unobserved factor variables.
Thus, individual elements of the vector would suffer from these omitted
variable biases. While this does not affect the estimation of the var-
iance, the decomposition components cannot be interpreted as accurate
causal estimates of the individual effect of a specific factor on test
scores. They only give an idea of their relative importance.
The estimation of the variance of the dependent variables is pre-
sented in the top part of Table 9. The individual, family and school
characteristics and the SDI are the set of factors. The values of 0.26,
0.24 and 0.22 show that only a quarter of all of the heterogeneity in the
scores of each area can be explained by the observed circumstances
beyond the students’ control. Consequently, the most important ex-
planatory factor is the difference in students’ efforts. The reliability of
this statistic is evidenced by the low value of the bootstrap standard
errors (1%), based on 100 replications.
The bottom part of Table 9 shows the Shapley-Shorrocks decom-
position of the variance. These results are presented both in levels and
as a percentage of the total variance. For the latter, the individual
circumstances account for 11.7%, 14.4% and 2.8% in math, reading
and science, respectively; the family circumstances account for 31.9%,
29.3% and 36.4%; and the school characteristics account for 44.8%,
43.1% and 46.0%.6 Social development alone accounts for 11.6%,
13.1% and 14.8%.
(6) To close, we decompose the school characteristics into their dif-
ferent contributions to have a better idea of the relative importance of
the effect of the SDI. It can be observed that social development ac-
counts for the variance in the test scores more than the combined effect
of the school size, the proportion of certified teachers, the shortage of
teachers in each subject, and whether the school is rural. Social de-
velopment is almost as important as school autonomy. This makes it
clear that local development has to be a key element in the design of a
broad education policy because it can contribute decisively to im-
proving academic achievement.
6. Conclusions
Improving educational quality is an important factor for develop-
ment, especially in Latin America, a region with large student perfor-
mance gaps which are a reflection of high income inequalities. This
study contributes to a broader understanding of the determinants of
6
This latter result is in line with the estimations of Treviño et al. (2010) and Castro
Aristizábal et al. (2018)) but contrasts with Cervini (2012), who found a smaller school
effect in a set of Latin American countries with data from SERCE.

91
G. Gimenez et al. Studies in Educational Evaluation 58 (2018) 80–96

educational achievement by considering the socio-economic environ-
ment of the area where the school is located. We test, for the specific
case of Costa Rica, whether the school district’s social development has
an effect on students’ test scores.
This research distinguishes itself from previous studies in the field in
some aspects. Most importantly, it combines PISA 2012 test scores with
a district-level Social Development Index (SDI), which is disaggregated
into its four components (health, economy, education and electoral
participation). This data set is used to estimate three-level hierarchical
models (students, schools and districts), which allow us to deal with
selection bias and unobserved heterogeneity problems. In addition,
while most of the literature about neighborhood effects takes the stu-
dent’s residence as the area of study, we consider the school’s location.
Results from the empirical analysis show, first, that individual and
family variables are more clearly linked with student achievement than
school variables, which is consistent with the PISA literature. Second,
that social development has a positive effect on test scores, which is
robust to the different disciplines considered —math, reading and sci-
ence. Our results also show that social development has positive but
diminishing effects on test scores.
The results are reinforced by the decomposition of the variance of
test scores in which the SDI’s contribution alone accounts for 11.6%,
13.1% and 14.8% in math, reading and science, respectively. These
effects are higher than the combined effect of the school size, the
proportion of certified teachers, the shortage of teachers in each sub-
ject, and the schools’ location.
The disaggregation of the SDI highlights that the economic and
education dimensions significantly and positively correlate with aca-
demic achievement. In contrast, a significant correlation of health has
not been found. Improving health conditions does not seem to make
much difference, as Costa Rica already has developed-country health
standards. With regard to electoral participation, the results are in-
conclusive.
Our findings have direct policy implications for the reduction of
differences in educational achievement, the increase of social cohesion
and the reduction of income inequality in the long term. Educational
policies should broaden their focus and not only target the school
—inputs such as school size, teachers or school autonomy, among
others—, but also the socio-economic characteristics of the environ-
ment where it is located. In particular, increasing household income
through better local job opportunities, expanding access to electricity
and ICTs, or investing in the infrastructure of the districts’ schools, may
be effective to boost academic test scores. Special emphasis should be
given to the lowest-ranked districts, where we have shown that ex-
pected returns from policies that increase social development are
highest. These districts tend to be far from the center of the country,
where San José is located, and generally receive less attention. Hence,
policy-makers should take this regional pattern into account.

Appendix A

Table A1
List of variables and sources.
Variable Definition Source Year

Student and family characteristics

Gender Gender of the student. Dummy variable: equals one if the student is male, zero otherwise PISA 2012
Age Age of the student in years. Continuous variable PISA 2012
Country of birth Country of birth. Dummy variable: equals one if the student was not born in Costa Rica, zero otherwise PISA 2012
Books at home Number of books available at home. Six-level categorical variable PISA 2012
Parents' occupation Index based on the highest occupational status of parents, which corresponds to the higher ISEI score of PISA 2012
either parent or to the only available parent’s ISEI score. Higher scores indicate higher levels of occupational
status. Continuous variable
Parents' education Parents' education expressed as years of schooling. Six-level categorical variable PISA 2012

School characteristics
Number of students Index of Number of students: total enrollment at school based on the enrolment data provided by the school PISA 2012
principal. Continuous variable
School ownership Refers to type of school. Dummy variable: equals one if the school is private, zero otherwise PISA 2012
Share of government funding Share of government funding. Continuous variable PISA 2012
School location Refers to the community in which the school is located, such as a village, hamlet or rural area (fewer than PISA 2012
3000 people), a small town (3000 to about 15,000 people), a town (15,000 to about 100,000 people), a city
(100,000 to about 1,000,000 people), close to the center of a city with over 1,000,000 people or elsewhere in
a city with over 1,000,000 people
Share of fully certified teachers at Share of fully certified teachers. Continuous variable PISA 2012
school
Shortage of math teachers Shortage of math teachers. Four-level categorical variable PISA 2012
Shortage of reading teachers Shortage of reading teachers. Four-level categorical variable PISA 2012
Shortage of science teachers Shortage of science teachers. Four-level categorical variable PISA 2012
School autonomy Index of school responsibility for resource allocation. Higher scores indicate higher levels of school PISA 2012
autonomy. Continuous variable

District-level characteristics
Social Development Index (SDI)
District-level index composed of 11 socioeconomic indexes. From 0 to 100. Higher scores indicate higher MIDEPLAN 2007-2012
levels of development. It proxies the possibilities of the population to access and enjoy a set of basic rights,
grouped in four dimensions: health, economy, education and electoral participation (see Section 3.1 and
Fig. 1 for an explanation).

92
G. Gimenez et al. Studies in Educational Evaluation 58 (2018) 80–96

Table A2
Summary statistics.
Variable Mean Sd Min Max

Average plausible values of test scores in

Math 405.974 62.794 195.210 694.742
Reading 440.636 66.570 196.378 660.552
Science 429.095 63.073 189.239 674.599

Student and family characteristics

Gender
Female 0.540 (Base category)
Male 0.465 0.499 0 1
Age 15.767 0.281 15.33 16.25
Country of birth
Costa Rica 0.967 (Base category)
Non-native student 0.033 0.179 0 1

Books at home
0-10 books 0.445 (Base category)
11-25 books 0.258 0.438 0 1
26-100 books 0.205 0.404 0 1
101-200 books 0.061 0.239 0 1
201-500 books 0.020 0.141 0 1
500 books or more 0.011 0.103 0 1
Parents' occupation 42.655 23.530 11.01 88.96

Parents' education
3 0.038 (Base category)
6 0.181 0.385 0 1
11 0.011 0.104 0 1
12 0.252 0.434 0 1
14 0.116 0.321 0 1
16 0.401 0.490 0 1

School characteristics
Number of students 855.910 614.925 26 4813
School ownership
Public 0.85 (Base category)
Privately operated 0.141 0.348 0 1
Share of government funding 77.631 31.283 0 100

School's community location

Village 0.233 (Base category)
Small town (< 3000) 0.265 0.441 0 1
Town (3000-150,000) 0.366 0.482 0 1
City (150,000–1,000,000) 0.113 0.317 0 1
Large city (> 1,000,000) 0.022 0.147 0 1
Share of fully certified teachers at 0.804 0.221 0 1
school

Shortage of math teachers

Not at all 0.635 (Base category)
Very little 0.273 0.446 0 1
To some 0.067 0.250 0 1
A lot 0.006 0.079 0 1

Shortage of reading teachers

Not at all (Base category)
Very little 0.260 0.439 0 1
To some 0.065 0.247 0 1
A lot 0.024 0.154 0 1

Shortage of science teachers

Not at all (Base category)
Very little 0.217 0.412 0 1
To some 0.125 0.331 0 1
A lot 0.011 0.105 0 1
School autonomy −0.666 0.836 −2.187 1.604

District-level characteristics
Social Development Index (SDI) 66.264 12.949 27.3 99.5
Health 76.598 12.391 50.8 98.1
Economy 38.500 17.047 6.3 100
Education 74.617 14.183 25.9 95.4
Electoral participation 43.292 9.017 18.1 70.2

93
 G. Gimenez et al.

Table A3
Correlation matrix of the Pisa scores, the Social Development Index and its categories.
Math Reading Science SDI Health Economy Education Electoral Births to Children with Infant Access to Consumption Access to Educational Enrollment in % of one- School
unmarried residential special teacher failure

Math 1
Reading 0.783 1
Science 0.849 0.783 1
SDI 0.301 0.306 0.313 1
Health 0.058 0.083 0.082 0.481 1
Economy 0.300 0.291 0.291 0.897 0.279 1
Education 0.292 0.305 0.315 0.865 0.214 0.684 1
Electoral 0.024 0.005 0.026 0.188 −0.196 0.046 0.125 1
participation
Births to 0.019 −0.008 0.025 0.276 0.388 0.161 0.094 0.273 1
unmarried
Children with low −0.024 −0.042 −0.016 −0.212 0.226 −0.334 −0.180 −0.142 0.015 1

94
weight
Infant 0.188 0.159 0.191 0.533 0.379 0.432 0.400 0.198 0.404 −0.032 1
mortality < 5
Access to 0.033 0.076 0.052 0.453 0.915 0.315 0.197 −0.285 0.161 −0.051 0.205 1
residential
Consumption of 0.219 0.213 0.206 0.779 0.230 0.926 0.551 −0.008 0.083 −0.336 0.317 0.293 1
electricity
Access to the 0.328 0.318 0.322 0.904 0.289 0.972 0.716 0.076 0.198 −0.308 0.472 0.305 0.811 1
Internet
Educational 0.166 0.164 0.186 0.515 −0.036 0.390 0.729 0.057 −0.008 −0.175 0.281 −0.024 0.327 0.400 1
infrastructure
Enrollment in 0.260 0.278 0.283 0.797 0.339 0.719 0.810 −0.110 0.072 −0.188 0.329 0.386 0.617 0.730 0.407 1
special
% of one-teacher 0.242 0.256 0.233 0.699 0.145 0.609 0.824 −0.070 −0.090 −0.148 0.232 0.200 0.525 0.617 0.602 0.698 1
School failure 0.100 0.109 0.121 0.277 0.053 0.084 0.319 0.496 0.243 0.015 0.274 −0.038 −0.046 0.158 0.069 0.045 −0.065 1

Notes: The development indices are expressed in relative terms (the higher the better). For example, in the case of under-5 mortality, the higher the value the better the relative situation of the district.
Studies in Educational Evaluation 58 (2018) 80–96
G. Gimenez et al.
References
Acar, E. (2011). Effects of social capital on academic success: A narrative synthesis.
Educational Research and Reviews, 6(6), 456–461. http://eric.ed.gov/?id=EJ936653.
Addison, T., Niño-Zarazúa, M., & Tarp, F. (2015). Aid, social policy and development.
Journal of International Development, 27(8), 1351–1365. http://dx.doi.org/10.1002/
jid.3187.
Aguilera, N., & Quintana, M. (2011). El peso de los niños y adolescentes y el rendimiento
escolar en México. El Trimestre Económico, 78(1), 115–141. http://www.redalyc.org/
articulo.oa?id=31340966004.
Ammermuller, A., Heijke, H., & Wößmann, L. (2005). Schooling quality in Eastern
Europe: Educational production during transition. Economics of Education Review,
(24), 579–599. http://dx.doi.org/10.1016/j.econedurev.2004.08.010.
Ayala, J., Marrugo, S., & Saray, B. (2011). Antecedentes familiares y rendimiento
académico en los colegios oficiales de Cartagena. Economía y Región, 5(2), 43–85.
Behtoui, A., & Neergaard, A. (2015). Social capital and the educational achievement of
young people in Sweden. British Journal of Sociology of Education, 37(7), 947–969.
http://dx.doi.org/10.1080/01425692.2015.1013086.
Bjørnskov, C. (2006). The multiple facets of social capital. European Journal of Political
Economy, 22(1), 22–40. http://dx.doi.org/10.1016/j.ejpoleco.2005.05.006.
Bourdieu, P. (1980). Le capital social. Actes de la Recherche en Science Sociales, 31, 2–3.
www.persee.fr/doc/arss_0335-5322_1980_num_31_1_2069.
Bronfenbrenner, U. (1977). Toward an experimental ecology of human-development.
American Psychologist, 32(7), 513–531. http://dx.doi.org/10.1037/0003-066X.32.7.
513.
Bruns, B., & Luque, J. (2014). Great teachers: How to raise student learning in latin America
and the Caribbean (advance edition). Washington, DC: The World Bank. License:
Creative Commons Attribution CC BY 3.0 IGO. http://www.worldbank.org/content/
dam/Worldbank/document/LAC/Great_Teachers-How_to_Raise_Student_Learning-
Barbara-Bruns-Advance%20Edition.pdf.
Burdick-Will, J., Ludwig, J., Raudenbush, S., Sampson, R., Sanbonmatsu, L., & Sharkey, P.
(2011). Converging evidence for neighborhood effects on children’s test scores: An
experimental, quasiexperimental, and observational comparison. In G. Duncan, & R.
Murnane (Eds.). Whither opportunity? Rising inequality, schools, and children’s life
chances (pp. 255–276). New York: Russell Sage.
Carlson, D., & Cowen, J. M. (2015). Student neighborhoods, schools, and test score
growth: Evidence from Milwaukee, Wisconsin. Sociology of Education, 88(1), 38–55.
http://dx.doi.org/10.1177/0038040714561801.
Case, A., Fertig, A., & Paxson, C. (2005). The lasting impact of childhood health and
circumstance. Journal of Health Economics, 24(2), 365–389. http://dx.doi.org/10.
1016/j.jhealeco.2004.09.008.
Castro Aristizábal, G., Gimenez, G., & Pérez Ximénez-de-Embún, D. (2018). Estimación de
los factores condicionantes de la adquisición de competencias académicas en América
Latina en presencia de endogenidad. CEPAL Review, 124, 35–59.
Castro Aristizábal, G., Gimenez, G., & Pérez Ximénez-de-Embún, D. (2017). Educational
inequalities in latin America, 2012 PISA: Causes of differences in school performance
between public and private schools. Revista de Educación, 376(April–Jun), 32–59.
http://dx.doi.org/10.4438/1988-592X-RE-2017-376-3433.
Cervini, R. A. (2012). El “efecto escuela” en países de América Latina: Reanalizando los
datos del SERCE. Archivos Analíticos de Políticas Educativas, 20(39), http://dx.doi.org/
10.14507/epaa.v20n.2012.
Chetty, R., Hendren, N., & Katz, L. (2016). The effects of exposure to better neighbor-
hoods on children: New evidence from the moving to opportunity experiment.
American Economic Review, 106(4), 855–902. http://dx.doi.org/10.1257/aer.
20150572.
Coleman, J. (1988). Social capital in the creation of human capital. American Journal of
Sociology, 94, 95–120. http://dx.doi.org/10.1086/228943.
Conn, K. M. (2014). Identifying effective education interventions in sub-Saharan Africa: A
meta-analysis of rigorous impact evaluations. Columbia University Academic
Commonshttp://dx.doi.org/10.7916/D898854G.
Currie, J., & Goodman, J. (2010). Parental socioeconomic status, child health, and human
capital. In D. J. Brewer, & P. J. McEwan (Eds.). Economics of education (pp. 156–162).
Amsterdam: Elsevier.
Dee, T. S. (2010). Education and civic engagement. In D. J. Brewer, & P. J. McEwan
(Eds.). Economics of education (pp. 89–92). Amsterdam: Elsevier.
Del Valle, R., & Fernández, A. (2014). Diferencias distritales en la distribución y calidad de
recursos en el sistema educativo costarricense y su impacto en los indicadores de
resultadosAvailable from San José, Costa Rica: Report prepared for the V Informe del
Estado de la Educación. (Retrieved 10 May, 2016) http://estadonacion.or.cr/files/
biblioteca_virtual/educacion/005/Andres_F_y_Roberto_Diferencias_distritales.pdf.
Durlauf, S. N. (2004). Neighborhood effects. In J. Henderson, & J. Thisse (Vol. Eds.),
Handbook of regional and urban economics: Vol. 4, (pp. 2173–2242). Amsterdam:
Elsevier. http://dx.doi.org/10.1016/S1574-0080(04)80007-5.
Durrant, G. B. (2009). Imputation methods for handling item-nonresponse in practice:
Methodological issues and recent debates. International Journal of Social Research
Methodology, 12(4), 293–304. http://dx.doi.org/10.1080/13645570802394003.
Evans, D. K., & Popova, A. (2015). What really works to improve learning in developing
countries? An analysis of divergent findings in systematic reviews. Policy research working
paper, No. WPS 7203. Washington, DC: The World Bank Group. (Retrieved May 16,
2016) http://documents.worldbank.org/curated/en/2015/02/24060240/really-
works-improve-learning-developing-countries-analysis-divergent-findings-
systematic-reviews.
Ferreira, F., & Gignoux, J. (2014). The measurement of educational inequality:
Achievement and opportunity. World Bank Economic Review, 28(2), 210–246. http://
dx.doi.org/10.1093/wber/lht004.
95
Studies in Educational Evaluation 58 (2018) 80–96
Freeman, R. B., & Viarengo, M. (2014). School and family effects on educational outcomes
across countries. Economic Policy, 29(79), 395–446. http://dx.doi.org/10.1111/1468-
0327.12033.
Galster, G. (2012). The mechanism(s) of neighbourhood effects: Theory, evidence, and
policy implications. In M. van Ham, D. Manley, L. Simpson, N. Bailey, & D.
Maclennan (Eds.). Neighbourhood effects research: New perspectives (pp. 23–56).
Dordrecht: Springer. http://dx.doi.org/10.1007/978-94-007-2309-2_2.
Giménez, G., & Castro, G. (2017). ¿Por qué los estudiantes de colegios públicos y privados
de Costa Rica obtienen distintos resultados académicos? Perfiles Latinoamericanos,
25(49), 195–223.
Grootaert, C. (2001). Does social capital help the poor? – A synthesis of findings from the local
level institutions studies in Bolivia, Burkina Faso, and Indonesia. Local level institutions
working paper No. 10. Washington, DC: The World Bank Group. (Retrieved July 16,
2016) http://documents.worldbank.org/curated/en/2001/06/1570743/social-
capital-help-poor-synthesis-findings-local-level-institutions-studies-bolivia-burkina-
faso-indonesia.
Grootaert, C., & van Bastelaer, T. (2001). Understanding and measuring social capital: A
synthesis of findings and recommendation from the social capital initiative. social capital
initiative working paper No. 24. Washington, DC: The World Bank Group. (Retrieved
July 16, 2016) http://siteresources.worldbank.org/INTSOCIALCAPITAL/Resources/
Social-Capital-Initiative-Working-Paper-Series/SCI-WPS-24.pdf.
Gundelach, B. (2016). Economic development and civic engagement in latin America: A
comparative study. Nonprofit & Voluntary Sector Quarterly, 45(2), 238–260. http://dx.
doi.org/10.1177/0899764015578287.
Hanushek, E. A. (2002). Publicly provided education. In A. J. Auerbach, & M. Feldstein
(Vol. Eds.), Handbook of public economics: Vol. 4, (pp. 2045–2141). Amsterdam:
North-Holland.
Hanushek, E. A., & Woessmann, L. (2012). Schooling, educational achievement, and the
latin American growth puzzle. Journal of Development Economics, 99(2), 497–512.
http://dx.doi.org/10.1016/j.jdeveco.2012.06.004.
Hanushek, E. A., & Woessmann, L. (2015). The knowledge capital of nations: Education and
the economics of growth. Cambridge, MA: MIT Presshttp://dx.doi.org/10.7551/
mitpress/9780262029179.001.0001.
Hanushek, E. A., & Wößmann, L. (2007). Education quality and economic growth.
Washington, D.C: World Bank. http://siteresources.worldbank.org/EDUCATION/
Resources/278200-1099079877269/547664-1099079934475/Edu_Quality_
Economic_Growth.pdf.
Hanushek, E. A., Link, S., & Woessmann, L. (2013). Does school autonomy make sense
everywhere? Panel estimates from PISA. Journal of Development Economics, 104,
212–232. http://dx.doi.org/10.1016/j.jdeveco.2012.08.002.
Harris, J. R. (2006). No two alike: Human nature and human individuality. New York: W.H.
Norton.
Harris, D. N. (2010). Education production functions: Concepts. In D. J. Brewer, & P. J.
McEwan (Eds.). Economics of education (pp. 127–131). Amsterdam: Elsevier.
Hernandez-Zavala, M., Patrinos, H. A., Sakellariou, C., & Shapiro, J. (2006). Quality of
schooling and quality of schools for indigenous students in Guatemala, Mexico and Peru.
Policy research working paper No. 3982. The World Bankhttp://dx.doi.org/10.1596/
1813-9450-3982.
Hillygus, D. S. (2005). The missing link: Exploring the relationship between higher
education and political engagement. Political Behavior, 27(1), 25–47. http://dx.doi.
org/10.1007/s11109-005-3075-8.
Jennings, P. A., & Greenberg, M. T. (2009). The prosocial classroom: Teacher social and
emotional competence in relation to student and classroom outcomes. Review of
Educational Research, 79(1), 491–525. http://dx.doi.org/10.3102/
0034654308325693.
Jensen, E. (2009). Teaching with poverty in mind: What being poor does to kids’ brains and
what schools can do about it. Alexandria: ASCD. http://www.ascd.org/publications/
books/109074.aspx.
Jiménez, M., Matus, J. A., & Martínez, M. A. (2014). Economic growth as a function of
human capital, internet and work. Applied Economics, 46(26), 3202–3210. http://dx.
doi.org/10.1080/00036846.2014.925079.
Klesner, J. L. (2007). Social capital and political participation in latin America: Evidence
from Argentina, Chile, Mexico, and Peru. Latin American Research Review, 42(2),
1–32. http://dx.doi.org/10.1353/lar.2007.0022.
Lee, J. W., & Barro, R. J. (2001). Schooling quality in a cross–section of countries.
Economica, 68(272), 465–488. Stable URL: http://www.jstor.org/stable/3549114.
Lupton, R. (2004). Schools in disadvantaged areas: Recognising context and raising quality.
Working paper, 76. London: Centre for Analysis of Social Exclusion. London School of
Economics. http://sticerd.lse.ac.uk/dps/case/cp/CASEpaper76.pdf.
Macdonald, K. (2014). PV: Stata module to perform estimation with plausible values. https://
EconPapers.repec.org/RePEc:boc:bocode:s456951.
Mander, A., & Clayton, D. (2007). Hotdeck: Stata module to impute missing values using the
hotdeck method. Available from: . (Retrieved June 24, 2016) http://econpapers.repec.
org/software/bocbocode/s366901.htm.
Masino, S., & Niño-Zarazúa, M. (2016). What works to improve the quality of student
learning in developing countries? International Journal of Educational Development, 48,
53–65. http://dx.doi.org/10.1016/j.ijedudev.2015.11.012.
McEwan, P. J. (2015). Improving learning in primary schools of developing countries: A
meta analysis of randomized experiments. Review of Educational Research, 85(3),
353–394. http://dx.doi.org/10.3102/0034654314553127.
Medina, F., & Galván, M. (2007). Imputación de datos: teoría y práctica. CEPAL: Serie es-
tudios estadísticos y prospectivos No. 54.. (Retrieved July 16, 2016) http://www.
cepal.org/es/publicaciones/4755-imputacion-datos-teoria-practica.
Meisenberg, G., & Woodley, M. A. (2013). Are cognitive differences between countries
diminishing? Evidence from TIMSS and PISA. Intelligence, 41(6), 808–816. http://dx.
doi.org/10.1016/j.intell.2013.03.009.
G. Gimenez et al.
Melguizo, A., & Perea, J. R. (2016). Mind the skills gap! Regional and industry patterns in
emerging economies. OECD development centre working paper No. 329. Paris: OECD
Publishinghttp://dx.doi.org/10.1787/5jm5hkp7v145-en.
MIDEPLAN (2013). Índice de Desarrollo Social 2013. Available from: San José, Costa Rica:
MIDEPLAN. (Retrieved May 20, 2016) http://www.mideplan.go.cr/component/
content/category/249:indice-de-desarrollo-social-2013.html.
Miguel, E., & Kremer, M. (2004). Worms: Identifying impacts on education and health in
the presence of treatment externalities. Econometrica, 72(1), 159–217. http://dx.doi.
org/10.1111/j.1468-0262.2004.00481.x.
Muennig, P. (2010). Education and health. In D. J. Brewer, & P. J. McEwan (Eds.).
Economics of education (pp. 80–88). Amsterdam: Elsevier.
Murillo, F. J., & Román, M. (2011). School infrastructure and resources do matter:
Analysis of the incidence of school resources on the performance of latin American
students. School Effectiveness and School Improvement, 22(1), 29–50. http://dx.doi.
org/10.1080/09243453.2010.543538.
Nahar, S. (2015). The concept of social development: Text analysis of the literature. Social
Development Issues, 37(1), 66–74. (9) http://www.ingentaconnect.com/content/icsd/
sdi/2015/00000037/00000001/art00007.
Nie, N. H., & Hillygus, D. S. (2001). Education and democratic citizenship: Explorations
into the effects of what happens in pursuit of the baccalaureate. In D. Ravitch, & J.
Viteritti (Eds.). Education and civil society. New Haven, CT: Yale University Press.
OECD (2003). Literacy skills for the world tomorrow–Further results from PISA 2000. Paris:
OECD Publishing. http://www.oecd.org/education/school/2960581.pdf.
OECD (2012). Untapped skills realising the potential of immigrant students. Paris: OECD
Publishing. http://www.oecd.org/education/school/Untapped%20Skills.pdf.
OECD (2013). PISA 2012 results: What makes schools successful? Resources, policies and
practices (volume IV), PISA. Paris: OECD Publishinghttp://dx.doi.org/10.1787/
9789264201156-en.
OECD (2014). PISA 2012 results: What students know and can do – Student performance in
mathematics, reading and science (volume I, revised edition). Paris: OECD
Publishinghttp://dx.doi.org/10.1787/9789264201118-en.
OECD (2015). The ABC of gender equality in education: Aptitude, behaviour, confidence.
Paris: OECD Publishing. https://www.oecd.org/pisa/keyfindings/pisa-2012-results-
gender-eng.pdf.
Owens, A. (2010). Neighborhoods and schools as competing and reinforcing contexts for
educational attainment. Sociology of Education, 83(4), 287–311. http://dx.doi.org/10.
1177/0038040710383519.
Ozier, O. (2014). Exploiting externalities to estimate the long-term effects of early childhood
deworming. Policy research working paper, No. 7052. Washington, DC: The World Bank
Grouphttp://dx.doi.org/10.1596/1813-9450-7052.
Parker, R. S. (2005). Putting social development to work for the poor: An OED review of world
bank activities. Washington, DC: The World Bank Group. (Retrieved July 16, 2016)
http://documents.worldbank.org/curated/en/2005/04/6432098/putting-social-
development-work-poor-oed-review-world-bank-activities.
Pelayo, J., & Brewer, D. J. (2010). Teacher quality in education production. In D. J.
Brewer, & P. J. McEwan (Eds.). Economics of education (pp. 178–182). Amsterdam:
Elsevier.
Rasbash, J., Leckie, G., Pillinger, R., & Jenkins, J. (2010). Children’s educational progress:
Partitioning family, school and area effects. Journal of the Royal Statistical Society:
Series A (Statistics in Society), 173(3), 657–682. http://dx.doi.org/10.1111/j.1467-
985X.2010.00642.x.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods. Thousand Oaks: Sage Publications.
Rothstein, R. (2010). Family environment in the production of schooling. In B. McGraw,
P. Peterson, & E. Baker (Eds.). International encyclopedia of education (pp. 148–155).
Amsterdam: Elsevier.
Rubin, D. B., & Schenker, N. (1986). Multiple imputation for interval estimation from
simple random samples with ignorable nonresponse. Journal of the American Statistical
Association, 81, 366–374. http://dx.doi.org/10.1080/01621459.1986.10478280.
Sahlgren, G. H. (2015). Real Finnish lessons: The true story of an education superpower4 Print
Surrey, UK: Centre for Policy Studies. (Retrieved July 16, 2016) http://www.cps.org.
uk/publications/reports/real-finnish-lessons-the-true-story-of-an-education-
superpower/.
Sampson, R. J., Morenoff, J. D., & Gannon-Rowley, T. (2002). Assessing “neighborhood
effects”: Social processes and new directions in research. Annual Review of Sociology,
96
Studies in Educational Evaluation 58 (2018) 80–96
28, 443–478. http://dx.doi.org/10.1146/annurev.soc.28.110601.141114.
Schuller, T. (2001). The complementary roles of human and social capital. Isuma:
Canadian Journal of Policy Research, 2(1), 18–24.
Schutz, G., Ursprung, H. W., & Wößmann, L. (2008). Education policy and equality of
opportunity. Kyklos, 61(2), 279–308. http://dx.doi.org/10.1111/j.1467-6435.2008.
00402.
Sen, A. (1999). Development as freedom. Oxford: Oxford University Press.
Shields, T. G., & Goidel, R. K. (1997). Participation rates, socioeconomic class biases, and
congressional elections: A cross-validation, 1958-1994. American Journal of Political
Science, 41, 683–691. http://dx.doi.org/10.2307/2111783.
Stock, J. H., & Yogo, M. (2005). Testing for weak instruments in linear IV regression. In D.
W. K. Andrews, & J. H. Stock (Eds.). Identification and inference for econometric models:
Essays in honor of Thomas Rothenberg (pp. 80–108). New York: Cambridge University
Press.
Teney, C., & Hanquinet, L. (2012). High political participation, high social capital? A
relational analysis of youth social capital and political participation. Social Science
Research, 41(5), 1213–1226. http://dx.doi.org/10.1016/j.ssresearch.2012.03.012.
Tian, M. (2006). A quantile regression analysis of family background factor effects on
mathematical achievement. Journal of Data Science, (4), 461–478.
Treviño, E., Valdés, H., Castro, M., Costilla, R., Pardo, C., & Donoso, F. (2010). Factores
asociados al logro cognitivo de los estudiantes de América Latina y el Caribe. Santiago de
Chile: UNESCO, LLECE http://www.unesco.org/new/es/santiago/resources/single-
publication/news/factores_asociados_al_logro_cognitivo_de_los_estudiantes_de/.
UNESCO (2014). Teaching and learning: Achieving quality for all. EFA global monitoring
report 2013/4Paris: UNESCO Publishing. (Retrieved July 16, 2016) http://en.unesco.
org/gem-report/report/2014/teaching-and-learning-achieving-quality-all.
United Nations Development Programme (2015). Human development index database.
(Retrieved June 20, 2016) http://hdr.undp.org/en/data.
Van Ham, M., & Manley, D. J. (2012). Neighborhood effects research at a crossroads. Ten
challenges for future research. Environment and Planning A, 44, 2787–2793. http://dx.
doi.org/10.1068/a45439.
Verba, S., Schlozman, L. K., & Brady, H. E. (1995). Voice and equality: Civic voluntarism in
American politics. Cambridge, MA: Harvard University Press.
Vigdor, J., & Ludwig, J. (2010). Neighborhoods and peers in the production of schooling.
In B. McGraw, P. Peterson, & E. Baker (Eds.). International encyclopedia of education
(pp. 163–169). Amsterdam: Elsevier.
Vivas Pacheco, H., Correa Fonnegra, J. B., & Domínguez Moreno, J. A. (2011). Potencial
de logro educativo, entorno socioeconómico y familiar: una aplicación empírica con
variables latentes para Colombia. Sociedad y Economía, 21, 99–124. http://www.
redalyc.org/articulo.oa?id=99622465007.
Wendelspiess Chavez Juarez, F., & Soloaga, I. (2014). IOP: Estimating ex-ante inequality
of opportunity. Stata Journal, 14(4), 830–846 StataCorp LP.
Wolfinger, R., & Rosenstone, S. J. (1980). Who votes? New Haven: Yale University Press.
Wooldridge, J. M. (1995). Score diagnostics for linear models estimated by two stage least
squares. In G. S. Maddala, T. N. Srinivasan, & Peter C. B. Phillips (Eds.). Advances in
econometrics and quantitative economics: Essays in honor of professor C.R. Rao (pp. 66–
87). Oxford: Blackwell.
World Bank (2005). Empowering people by transforming institutions: Social development in
world bank operations. Washington, DC: The World Bank Group. (Retrieved July 16,
2016) http://documents.worldbank.org/curated/en/2005/01/5583580/
empowering-people-transforming-institutions-social-development-world-bank-
operations.
World Bank (2012). ICTs for greater development impact: World bank group strategy for in-
formation and communication technology 2012–2015. Available from Washington, DC:
World Bank. (Retrieved September 20, 2016) http://siteresources.worldbank.org/
EXTINFORMATIONANDCOMMUNICATIONANDTECHNOLOGIES/Resources/WBG_
ICT_Strategy-2012.pdf.
Wößmann, L. (2005). Educational production in Europa. Economic Policy, 20(43),
445–504. https://ideas.repec.org/a/bla/ecpoli/v20y2005i43p445-504.html.
Wößmann, L. (2003a). Educational production in East Asia: The impact of family background
and schooling policies on student performance, working paper No. 1152. Kiel: Institute for
World Economics. http://en.agi.or.jp/workingpapers/WP2003-17.pdf.
Wößmann, L. (2003b). Schooling resources, educational institutions and student perfor-
mance: The international evidence. Oxford Bulletin of Economics and Statistics, 65(2),
117–170. http://dx.doi.org/10.1111/1468-0084.00045/pdf.