See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/356906033

Zones of Educational Exclusion of Out-of-School Youth

Article  in  International Journal of Educational Development · December 2021

DOI: 10.1016/j.ijedudev.2021.102532

CITATION READS

1 199

2 authors:

Marcos Delprato Alessia Frola

University of Sussex University of Sussex
40 PUBLICATIONS   316 CITATIONS    2 PUBLICATIONS   1 CITATION   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Spatial Education Inequality (SEI) View project

Monitoring Global Education View project

All content following this page was uploaded by Marcos Delprato on 09 December 2021.

The user has requested enhancement of the downloaded file.

 Zones of Educational Exclusion of Out-of-School Youth

Marcos Delprato*
Centre for International Education, University of Sussex, Brighton BN1 9QQ UK
Email: [email protected]

Alessia Frola
Centre for International Education, University of Sussex, Brighton BN1 9QQ UK
Email: [email protected]

Abstract

This study provides new evidence on bottlenecks of progression across six zones of
exclusion from primary to secondary education for out-of-school youth of five poorer
countries (Guatemala, Honduras, Panama, Paraguay and Senegal) using the recent out-
of-school assessment from PISA-D. We also investigate whether these barriers have
gendered dimensions and which policies can boost progression. We find that barriers
have differential impacts across exclusion zones, with low parental education and
involvement in education and local language being leading determinants, and some of
these barriers having a larger impact for girls. We also find that policies linking
financial support to attendance, raising teaching quality and tackling discrimination
push up youth cohorts to top zones of exclusion, thereby allowing them to gain more
education.

Keywords: zones of exclusion; SDG4; SDG5; PISA for development; Out-of-school youth.

* Corresponding author.

1
1. Introduction

Access to education improved considerably in the decade following 2000, and as Lewin (2011)

suggests, it “lies at the very heart of development” (p.8) for its multiple positive effects on the

economy (Orozco and Validivia, 2017), poverty relief (Lewin, 2011), social struggle

(Kameyama, 2018), personal and community empowerment (Tarallo, 2019). However, recent

estimates indicate that there are currently 62 million adolescents of lower secondary age out-

of-school (nearly a quarter of the total 258 million children, adolescent and youth who were

out-of-school in 2018), whose majority is located in middle- and low-income countries (Ward,

2020). It is also worrying that recent progress in access has stalled, with the combined number

of out-of-school children and youth only declining by just above 1 million per year since 2015

(UIS, 2019).

At a global level, the current COVID-19 pandemic resulted in visible learning losses,

holding back some of the progress made so far on access to education and narrowing learning

gaps. In fact, the emphasis on equity within SDG4 (Target 4.5: “By 2030, eliminate gender

disparities in education and ensure equal access to all levels of education and vocational

training for the vulnerable, including persons with disabilities, indigenous peoples and children

in vulnerable situations”) is likely to become futile if not addressing ongoing issues behind

access. Certainly, making the fulfilment of SDG4 a reality and fully succeeding at granting

universal access to education, rely on enacting strategies aimed at promoting not solely the

“right to education” (universal access), but working toward the enhancement of rights in

education (Kameyama, 2018) by increasing school retention and attendance regardless of

social status, gender, geographical location, religion and social identity.

The empirical literature suggests that the most striking educational inequalities lay at the

intersection between gender, Indigeneity1 and rurality, ultimately linked with poverty and

economic struggle (Richardson et al., 2019). The multilayered and stratified nature of

2
educational inequalities behind participation and retention of out-of-school youth (henceforth

denoted as OOSY), whose components “reinforce and exacerbate each other” (Kabeer and

Santos, 2017, p.6), and the need for a locally specific and detailed plan of interventions, all

need to be supported by updated data from out-of-school populations and their learning

experiences. We contribute to this debate by presenting novel empirical evidence concerning

the drivers behind the educational exclusion of youth for a range of poorer countries.

In particular, based on the out-of-school assessment from PISA for development (PISA-

D), the paper’s overall aim is to assess which are the leading determinants on access and

barriers across the different stages of progression (from primary to secondary education)

among OOSY in five low- and middle-income countries: Guatemala, Honduras, Panama,

Paraguay, and Senegal. Given the discrete nature of the educational zones of exclusion (the

dependent variable), ordered logit models are used to answer the following three research

questions:

1. Which are the most powerful factors determining where students fall among

educational zones?

2. Do barriers for progression across the different zones such as language spoken,

different work types and lack of parent’s educational input affect girls more than boys?

3. Among the youth who entered education systems, what is the role of school factors and

the leverage effects of diverse kind of financial support for increasing attendance, and

what are the conditions under which students may continue schooling?

The paper is structured as follows. Section 2 contains a review and contextual background,

and Section 3 describes the data used. Section 4 outlines the methods employed, and results

are presented in Section 5. We offer some conclusions and policy implications in Section 6.

3
2. Background

2.1. Definition and approaches

There are many definitions of out-of-school populations; Vayachuta et al. (2015) define it “as

the population between the ages of 3-25 years who do not attend schools or who have studied

in school but dropped out before completing their compulsory or basic education” (p.1).

However, such definition alone fails at acknowledging the complexity of the phenomenon. It

is vital to grasp the disengagement process behind dropout and educational (and social)

exclusion (Fortin, et al. 2006; Silver, 2007; Peters, 2010; Ananga, 2011; Bowers and Sprott,

2012; Frazelle and Nagel, 2015; UNICEF, 2015) as well as the multi-layered disadvantages

and inequalities behind the phenomenon (Kabeer and Santos, 2017; Kameyama, 2018).

The literature offers various models and frameworks which holistically approach the issue

of OOSY. To comprehensively analyse various forms of disadvantages, Singh and Mukherjee

(2018) offer a representation of environmental systems affecting individuals’ trajectories,

focusing on temporality and process. They refer to these environments as the microsystem (an

individual’s immediate surroundings: family, school, religious institutions, clubs, peer group);

the macrosystem (laws, policies, belief system, socioeconomic status, economic patterns and

social conditions); and finally, the chronosystem -i.e., the interplay of events during an

individual’s life-course (Singh and Mukherjee, 2018). Similarly, Behrman et al. (2014)

attribute the roots of such phenomenon to various and interconnected reasons which focus on

individuals and their immediate surroundings (household/family), their communities and the

macroenvironment (the larger political, social, and economic structure). When addressing the

origins of educational exclusion, specifically in Sub-Saharan Africa (SSA), Lehman (2003)

offers a multidimensional approach that tackles three kinds of distance: a physical one (distance

to school and the quality of the infrastructure connecting the individuals to schools); a temporal

one (actual time to go to school, found negatively correlated with school performance and
4
attendance (Pôle de Dakar and UNESCO BREDA, 2005); and, finally, the cultural distance

which is “the closeness of societal views about the value and benefits of education to society”

(Lehman in Daly et al., 2016, p. 183). The study of Daly et al. (2016) on participation and

retention in the Gambia adds a fourth dimension to the model: financial distance, concerning

household and school economy and finance (Hannun and Adams, 2007).

The OOSY group is a highly heterogeneous group whose characteristics vary significantly

among regions and communities. UNICEF and UIS divide OOSY into two main groups

needing specific and targeted interventions: (i) those “presently out-of-school”, primary and

secondary age children not enrolled for more than a year; (ii) children and adolescents

considered “at risk of being out-of-school”, such as overage students and those attending

irregularly (Lewin, 2013; Kameyama, 2018).

Trends regarding OOSY vary among countries and across communities, calling for locally-

based analysis of heterogeneous set of interventions (Lewin, 2013). For example, in Senegal,

most OOSY will never attend school (UNESCO, 2014), while countries from Central and

South America face enormous challenges due to high dropout rates (Aldeman and Székely,

2017). These patterns can be identified and then adequately addressed following Lewin’s

(2007) approach, which broadly conceives the idea of access to education by identifying six

different “zones of exclusion”. Such approach, explained in Section 3.1, goes beyond “simple

definitions of enrolment registrations (and) need to be accompanied by a broader vision which

includes entry level and progression at an appropriate age, regular attendance, satisfactory

achievement, appropriate chances to progress to post-primary, and more equitable distribution

of opportunity” (Lewin, 2009, p. 171).

2.2. Major factors of exclusion

There are many reasons for educational exclusion which are cumulative and do not always

revolve around economic issues (Kabeer and Santos, 2017). The following paragraphs outline
5
some of the most significant barriers to education (e.g., poverty, gender and Indigeneity) which,

combined, negatively exacerbate their effects on access to education.

To begin with, household wealth is an accurate predictor of educational access, retention

and attainment; therefore, poverty (and its circular repercussions) is a leading determinant for

the lack of education and social outcomes (Vayachuta et al., 2015; Orozco and Valdivia, 2017;

Aldeman et al., 2018; Richardson et al., 2019). The opportunity costs of education are

considerably high in middle- and low-income countries where the immediate economic benefit

of child and youth work might be preferred by the household and the individual, in contrast to

education and its long-term economic benefits (Aldeman and Székely, 2017).2 However,

interesting evidence from Honduras shows that economic reasons lay behind most males

dropouts, while females primary reasons for dropout are “personal reasons, as having to

perform households tasks” (Aldeman and Székely, 2017, p. 246). It is vital to disentangle such

“personal reasons” and evaluate all those additional variables competing at endangering

females’ education. Tarallo (2019) rightfully writes that “gender-based educational inequality

is a worldwide epidemic” (p.4).

When it comes to access to education, although gender gap has been reduced, many girls

and female adolescents are at high risk of dropout. Keeping girls in school would not solely

benefit the individual but also society as a whole. From a strictly economic and developmental

perspective, communities and states lose a large amount of potential productivity because of

women’s inactivity (Chaaban and Cunningham, 2011). Furthermore, following a right-based

approach, societies have the moral duty to grant women access and a successful transition

within educational systems (Kameyama, 2018).

One of the most significant barriers to girls and female adolescents’ participation is related

to the lack of proper school facilities and infrastructures. In rural areas, where a significant

number of disadvantaged females live, generally lack basic hygienic facilities, transportation

6
and markets (Aldeman and Székely, 2017; Tarallo, 2019). Such inefficiencies affect teachers

in general as well. The lack of (quality) teachers highly impacts the quality and value of the

education received (Mulkeen and Chen, 2008; Daly et al., 2016), which, coupled with the lack

of schools (especially at the secondary level), discourage children and adolescents from

continuing education. Additional reasons behind dropouts are household and caring labour,

safety issues related to violence and the emotional distress if a child/youth has to move away

in order to attend school (Menjívar, 2008; Guinan, 2015; Hajny, 2015; Davis, 2016; Orozco

and Valdivia, 2017; Seeley, 2018).

In addition to the barriers described above, early marriage and teen pregnancy constitute

some of the biggest challenges for educational participation, widening education gender gaps.

Again, trends are reportedly heterogeneous among regions. Lloyd and Mensch (2008) explain

how early marriage rather than pregnancy negatively affects girls’ participation in the SSA

region (see also: Delprato et al., 2017, 2015), while Orenstein (2017) defines early pregnancy

as “an epidemic” in Latin America. Early marriage through childbirth causes health

complications (Delprato and Akyeampong, 2017), it is positively associated with dropout

(Dixon-Mueller, 2008), negatively impacts on future economic opportunities (Arceo-Gómez

and Campos-Vázquez, 2014), and highly discourages school access and participation (Murphy-

Graham et al., 2020). More broadly, gender-based violence and patriarchal societies (Menjívar,

2008; Guinan, 2015), as well as little political representation (Ogrodnik and Borzutzky, 2011;

Tarallo, 2019), are significant barriers for female school access and retention. Nevertheless,

educated women can increase their children’s possibilities to attend school and transition from

one level to another successfully. For instance, Hendrick and Marteleto (2017) show how

Honduran mothers’ decision-making ability within the household has positive outcomes on

adolescents’ educational attainments and aspirations, demonstrating the importance of the

immediate familiar environment and female role models (Johnston et al., 2014).

7
 Indigenous youth face numerous disadvantages and struggle for the right to education. They

face social exclusion, and their status was shown to be negatively correlated to school

enrolment and positively affecting dropout (Aldeman and Székely, 2017). Besides, Indigenous

populations are among the poorest and most disadvantaged (Tarallo, 2019; Nichols, 2021).

Ultimately, Indigenous people’s educational exclusion “revolves around the phenomenon of

social discrimination”, i.e., upon their social identity (Kabeer and Santos, 2017, p.6). Social

discrimination is reflected by a unified school curricula, the denial of special education, and a

neglect of linguistic barriers.

Nichols (2021) points at institutional racism in Guatemala as the most significant form of

exclusion and discrimination for Indigenous youth. Batz (2018) advocates for the

decolonisation of knowledge, creating specific curricula which would increase Indigenous

students’ performances, inclusion and motivation. The epistemological dilemma regarding

Indigenous education is reflected by the absence of an adequate number of Indigenous teachers

and intercultural education (Nichols, 2021), and the lack of specific school material (Grasso,

2018). Also, most Indigenous people live in rural areas that offer hardly any adequate

transportation and cannot grant comprehensive financial support (Vayachuta et al., 2015;

Grasso, 2018). Hence, unsurprisingly, exclusively economic interventions aimed at increasing

access and retention in education have had little to no impact. Interventions need to surpass

deep challenges such as the lack of context-specific curricula and linguistic barriers (i.e., not

speaking the language of instruction at home, being used as a persistent discrimination

mechanism negating Indigenous people’s rights to access and to transmit their culture through

it),3 with a significant negative impact on access and performance (Grasso, 2018; Nichols,

2021; Richardson et al., 2019).

8
3. Data

The analysis relies on the out-of-school assessment of PISA for Development (PISA-D),

comprising OOSY aged 14-16 years old using a household-based survey.4 The OOSY survey

is a spin-off from the main PISA-D school-based assessment where five out of the nine

countries that participated in PISA-D undertook the out-of-school component: Guatemala,

Honduras, Panama, Paraguay and Senegal. Hence, the OOSY is one of the three strands5 of

PISA-D, which, ultimately, aims to make PISA more accessible and relevant to both low and

middle-income countries following the educational prosperity framework (Willms, 2018).

PISA-D is designed to portray the reality of resource-constrained education systems with more

significant populations of marginalised/poor students (OECD, 2018a, 2018b). This is

accomplished by using different test instruments and gathering contextual information about

specific bottlenecks of schooling processes in the global south. For example, PISA-D offers a

wider coverage at the lower end achievement scales (where 60% of the testing items are used

to measure performance at level 2 or below). Also, contextual questionnaires contain new

information on factors halting students' progress, such as child labour, travel time to school,

food security, health, and specific education policies (Ward, 2018).

The OOYS component offers a tool to identify the most disadvantaged within education

systems: those who left or are at risk of leaving education (alongside an assessment of the

reasons for exclusion). This phenomenon depicts a bleak scenario where around 37%-58%

(7%-23%) are out-of-school at the lower and upper-secondary in the SSA region (LAC region)

(see: UNESCO, 2020a). For the construction of the OOYS sample across the five countries, a

sample of 89,000 households was drawn and, after the application of the probability and non-

probability approaches, more than 7,200 respondents aged 14-16 (who were either not in school

or were in school in grade six or below) completed the household-based assessment (for more

details on the OOYS and assessment, see: Ward, 2020).

9
3.1 PISA-D and OOSY limitations

Relying solely on PISA-D for the empirical analysis does have limitations (related to the data’s

coverage and how the information is gathered). One of the fundamental issues is OCED’s

implicit homogenous view on education systems and an ideal culture-free assessment (Bray et

al., 2020) relying on an economistic approach to education, emphasising a specific kind of

datafication and accountability (e.g., Addey, 2017, 2020; Fischman et al., 2018). In turn, some

studies (e.g., Meyer and Benavot, 2013; Tröhler, 2013) argue on the unfeasibility of a cultural

neutral educational platform, thereby claiming that OECD learning assessment are

inadequately sensitive to cultural differences.

A related additional issue is language –and the translation and adaption of surveys

instruments– raising concerns on whether the data obtained is fully comparable and reliable

(Asil and Brown, 2016; Bray et al., 2020). Addey and Gorur (2020) analyse the process of

PISA-D translation of instruments (technical, political, semiotic elements), concluding: “the

making of PISA-D was not just a technical exercise of translating survey instruments of PISA

and making them ‘fit’ for low-income nations; the new low-income contexts were also

translated and made to fit PISA –and this political exercise had to go hand in hand with the

technical one– without the political part, the technical could not succeed.” (p. 549)6

Moreover, although PISA-D is conceived to depict the reality of students in poor education

systems (Ward, 18; Willms, 18), claiming that the survey has been developed in partnership

with low- and middle-income countries (Ward and Zoido, 2015), a contrasting view has been

put forward (Auld et al. 2019; Xiaomin and Auld, 2020) with local agency and consensus

lacking in the application of PISA-D in Cambodia (one of the seven PISA-D countries

participating) as well as a lack of concerns about the utility of it for the country (Xiaomin,

2021). However, in the case of Ecuador and Paraguay, their participation was self-funded and

guided by the “global ritual of belonging” to international learning assessments showing a

10
greater commitment beyond donors’ influences with different objectives such as international

accountability and transparency in Paraguay or acquiring technical support in Ecuador (Addey,

2020). Auld et al. (2020), after analysing PISA-D country reports’ findings and

recommendations, indicate that, though PISA-D made novel technical adjustments, it lacks

criticality given the weak emphasis on addressing the challenges of these education systems.

The OOSY, being a spin-off survey of PISA-D, has the same earlier limitations highlighted

when used to assess key issues behind lack of youth progress throughout the education life

course. Also, the OOSY component has further limitations. In a recent blog,7 Xiaomin (2021)

draws attention to two key concerns on its design, namely: (i) the conceptual definition of

OOSY which lacks consistency as it includes two groups (i.e., repeaters and those who are not

attending school regularly; in average 27% of the total sample) who are actually in school; (ii)

the lack of purposed sampling following the recommendations (Carr-Hill, 2015) that the

majority of marginalised OOSY lives in either urban slums or isolated disadvantaged rural

areas. The author concludes that a redefinition of the concept of out-of-school is needed, so as

to enhance PISA-D experimental validity as a tool to bring OOSY populations into the

monitoring framework of SDG4.

The implications of the conceptual and technical details of the construction of the OOSY

sample for the empirical analysis are reflected in various limitations. On the one hand,

conceptually, the homogeneity of the survey does not allow us to pin-down cultural differences

between the five countries included in the sample, and issues are analysed under the OECD

economic view of education systems without full involvement of countries in its design. On

the other hand, technically, the main drawback is the OOSY’s small sample sizes and related

lack of precision for estimates, particularly given the absence of over-sampling for those groups

more likely to be out of school (e.g., youth living in either poor isolated areas or large urban

11
deprived areas). The findings of the paper need to be understood within the context of these

limitations.

3.2 Zones of educational exclusion

The final working sample consists of 7,887 observations for five countries: Guatemala,

Honduras, Panama, Paraguay and Senegal.8 Our definition of the dependent variable measures

different stages of schooling, from primary to secondary, according to zones of exclusion in

line with the CREATE (Lewin, 2011, 2017) and UNICEF out-of-school-children initiatives

(UNICEF, 2018). This framework captures the nuanced causes and subtleties behind each zone

where patterns of exclusions are likely to be different, as well as the strategies that address

them. Our approach follows Ward’s (2020) six zones of exclusion: zone 1 (never enrolled),

zone 2 (primary dropout), zone 3 (enrolled but at grade 6 or below), zone 4 (leave at grade 6,

primary leavers), zone 5 (dropout at secondary) and zone 6 (fading out, enrolled but overaged).9

Table 1 (Panel A) shows the distribution of the discrete dependent variable. On average,

across the five countries, nearly half of the 14–16-year-old youth tends to leave school at the

end of primary school (24%, zone 4) or drop out at the early grades of secondary school (21%,

zone 5). There is also a significant population at risk of dropping out (zone 3; with a rate of

22%), and more than 10% of youth in the sample never entered the education system at all

(zone 1, 13%).10

[Table 1 here]

3.3 Covariates

The set of explanatory variables employed is shown in Panel B (Table 1). Youth and family

factors tend to worsen, with established drivers of disadvantage rising when moving (in reverse

order) across exclusion zones. Consider, for instance, youth characteristics. Table 1 shows that,

while 40%-49% do not speak the language of the test11 (columns 2 and 3, primary level-dropout

12
or at risk of doing so), this rate goes down to 32% for secondary dropouts (column 5). Equally,

the prevalence of paid or family work is 10% lower when comparing secondary to primary

dropouts (column 5 versus column 2), as well as youth’s welfare being lower for primary

dropouts (index of zero versus -.0.19). Likewise, family disadvantages are more pronounced at

earlier zones, and mechanisms of education transmission are weaker for youth who do not

make it into secondary education. For instance, primary dropouts are more than double

compared to secondary dropouts if a youth comes from a poor household (66% versus 29%),

and mother’s literacy gap among these two groups is also more than double (of 0.26 for primary

dropouts and 0.60 for secondary dropouts) and primary youth leavers are more prone to live in

less safe areas surrounding schools.

4. Empirical framework

The model we rely on is an ordered logit (OL) as the dependent variable is discrete. We focus

on variations regarding predictions as we are interested in modelling the impact of different

covariates across the different zones youth/students fall.

4.1. Modelling zones of exclusion key determinants and gender interactions

For the first two research questions we run an OL based on a dependent variable taking 6

values: y = 1 (zone 1), y = 2 (zone 2),…, y = 6 (zone 6) as defined in section 3.1. Formally, the

latent variable model is: 𝑦𝑖∗ = 𝑋𝑖 𝛽 + 𝜀𝑖 , where 𝑋𝑖 are the covariates (Table 1, Panel B) for

individual i. The measurement of the discrete outcome is divided into J (=6) ordinal categories:

𝑦𝑖 = 𝑚 if 𝜏𝑚−1 ≤ 𝑦𝑖∗ < 𝜏𝑚 for 𝑚 = 1 to 𝐽. The probability of observing the dependent

variable is the difference on cumulative probabilities for two adjacent categories:

Pr(𝑦 = 𝑚 |𝑋) = 𝐹 (𝜏𝑚 − 𝑋𝛽 ) − 𝐹 (𝜏𝑚−1 − 𝑋𝛽 ) (1)

where F is the logistic cumulative distribution function for 𝜀 with variance Var(𝜀 ) = 𝜋 2 /3.

13
 Also, to gauge whether there is a differential impact of barriers by gender, we run the same

model but interacting gender (G) dummy with work indicators (W), Indigeneity (I) (language

spoken at home), and parental education (E) variables. So, the latent variable model, where

probabilities are derived, is:

𝑦𝑖∗ = 𝑋𝑖 𝛽 + 𝛾1 (𝐺𝑖 x 𝑊𝑖 ) + 𝛾2 (𝐺𝑖 x 𝐼𝑖 ) + 𝛾3 (𝐺𝑖 x 𝐸𝑖 ) + 𝜀𝑖 (2)

Based on this model, we obtain predicted probabilities for each of the six zones for

combinations of gender with: (i) Indigeneity, mother literacy, finish education; and (ii) work

indicators (paid, care, chores, and family/farm). For example, for the first set of combinations,

the estimated probabilities for category m (m=1,…,6):

̂(𝑦 = 𝑚|𝑋; 𝐺 = 1, 𝐼 = 1, 𝑀𝐸𝑑 = 0, 𝐹𝐸 = 0) = 𝐹(𝜏̂ 𝑚 − 𝑋𝛽̂ ) − 𝐹(𝜏̂ 𝑚−1 − 𝑋𝛽̂ )

Pr (3a)

̂(𝑦 = 𝑚|𝑋; 𝐺 = 1, 𝐼 = 0, 𝑀𝐸𝑑 = 1, 𝐹𝐸 = 1) = 𝐹(𝜏̂ 𝑚 − 𝑋𝛽̂ ) − 𝐹(𝜏̂ 𝑚−1 − 𝑋𝛽̂ )

Pr (3b)

̂(𝑦 = 𝑚|𝑋; 𝐺 = 0, 𝐼 = 1, 𝑀𝐸𝑑 = 0, 𝐹𝐸 = 0) = 𝐹(𝜏̂ 𝑚 − 𝑋𝛽̂ ) − 𝐹(𝜏̂ 𝑚−1 − 𝑋𝛽̂ )

Pr (3c)

̂(𝑦 = 𝑚|𝑋; 𝐺 = 0, 𝐼 = 0, 𝑀𝐸𝑑 = 1, 𝐹𝐸 = 1) = 𝐹(𝜏̂ 𝑚 − 𝑋𝛽̂ ) − 𝐹(𝜏̂ 𝑚−1 − 𝑋𝛽̂ )

Pr (3d)

A comparison of estimated probabilities based on Eq. (3a)-(3b) allows us to assess the

profile of exclusion between Indigenous girls (I=1, G=1) having a literate mother (MEd=1) and

a family which supports resuming education (FE=1), against Indigenous girls not having a

literate mother (MEd=0) and also lacking family support for finishing education (FE=0).

Likewise, Eq. (3c)-(3d) denote the same profiles for the likelihood of falling into the different

zones of exclusions for boys (G=0).

4.2. Model for “ever attended” youth’s sample and conditions for progression

Regarding the third research question, which is based on the youth sample who ever attended

school, we also run an OL with the dependent variable taking 4 values: from 1 (zone 2) to 4

(zones 5 and 6). Within this model, we obtain marginal effects (MEs) for three sets of policy

14
covariates: school factors (SF), financial support for attendance (FS), and conditions for

continuing in schools (CC).

More specifically, we estimate the discrete change on the probability m for a change on

𝑋𝑘 = 0 to 𝑋𝑘 = 1. For instance, the discrete change due to financial support (lack of FS to the

presence of FS) is calculated as:

∆Pr (𝑦 = 𝑚|𝑋)
= Pr(𝑦 = 𝑚|𝑋, 𝐹𝑆𝑘 = 1) − Pr(𝑦 = 𝑚|𝑋, 𝐹𝑆𝑘 = 0) (4)
∆𝐹𝑆𝑘 (0 → 1)

And, similarly, for the other MEs based on the SF and CC set of covariates.

5. Results

5.1 Research question 1 – determinants of youth falling into educational zones

Results for the impact of youth characteristics and their families’ characteristics on zones of

exclusion are shown in Table 2. OL estimates are presented as odds-ratio (OR) in a stepwise

manner; in the first model, students/youth variables are added and then, in the second model,

we insert family covariates (in both cases, we include dummies to account for country

heterogeneity in the pooled sample).12 After testing these two specifications, the fuller model

proved to be a better fit compared to the model with only students/youth covariates.13 Table 2

also displays the percentage change in odds by standard deviation increase (%StdX) for

continuous variables, alongside estimated OR.

[Table 2 here]

Table 2 (columns 2 or 5 and 6) displays the array of barriers faced by youth from poorer

countries to move across stages at primary and secondary levels. On the one hand, among

youth’s characteristics, prominent barriers are constituted by different kinds of work, albeit

their impacts are not homogenous. Performing caring activities or working for the family

business/farm has negative impacts on the chances of moving across zones (from zones 1 to 2

15
to 3, and so on) of -14.9% and -21.5%, while other working activities (paid work and chores)

enhance access and progression by 7.7 and 32.6 per cent (see column 5). The odds of moving

across educational zones are 5.3% higher for girls than boys. Perhaps surprisingly, non-native

speakers of the test language are more likely to move up educational stages by 41%, while the

impact of disability is not statistically significant. Also, good health and individual well-being

contribute to progression: by the increment of 1Std in the indices for the odds of moving-up

zones increase by 4.2%-6.5% (column 6).

On the other hand, among family factors, being from a poor household is related to 20.5%

fewer chances to move across zones; estimates also indicate the importance of education

transmission from parents to their children –i.e., youth whose mothers and fathers are literate

have around 65% and 45% more chances to move across zones. Also, note the leading effect

of the “emphasis placed on education by parents” on the odds of youth’s progression (OR =

2.02), which is likely to reflect parents' interaction and productivity in their children's

education. Living in a safe area (home-school) is associated with 14% more chances to progress

across zones (column 6). With regards to countries’ ranking, youth from Senegal have the

lowest chances to progress across zones (OR=0.55), followed by Guatemala (OR=1), Honduras

(OR=1.79), Paraguay (OR=2.16) and Panama (OR=2.34) (see column 2).

We can better understand educational progression’s bottlenecks by plotting the marginal

effects (MEs) for each of the six zones of exclusion by background characteristics. We plot

these in Figure 1 for selected covariates. To begin with, consider parental literacy. Compared

with those youths whose parents are illiterate, we find that having either a literate mother or

father increases the probability of reaching zones 4 or 5 (primary leavers or secondary dropout)

by 0.03-0.05 but, conversely, these covariates have adverse effects for zones 1 or 2 (never

enrolled or primary dropout/first grades) of around 0.02-0.07. The same pattern of associations

can be seen for the covariate “family places importance on finishing education”, with positive

16
MEs of 0.04-0.08 for zones 5 and 6 but negative impacts of -0.05 (zone 2) and -0.10 (zone 1).

In contrast to those from wealthier households, youth from poor households are 2% and 4%

more likely to never have attended school and drop out during the early grades of primary

(zones 1 and 2); this pattern is reserved (becoming negative) for zones 4 to 6.

[Figure 1 here]

Moreover, not speaking the test language at home (or being Indigenous) versus speaking it

becomes a strong determinant for youth falling into the bottom two zones of exclusion, with

negative MEs of 0.02 and 0.05. Interestingly (probably due to a positive selection/resilience),

Indigeneity is associated with positive and significant higher probabilities (of around 3% and

4%) of youth who actually make it to the end of primary and early grades of secondary. We

do not find notable differences for the MEs of gender, disability or paid work across zones.

5.2 Research question 2 – the differential impact of barriers by gender

Here, we investigate whether gender effects are compounded with further disadvantages for

youth progression across zones.14 Table 3 displays OR-based estimates that rely on the model

with interactions (Eq. (2)). Estimates for the interaction terms of gender with other key

covariates indicate a gendered entrenchment of inequalities within the OOYS population for

the five countries underpinned by widening disadvantages for girls compared to boys.

[Table 3 here]

Consider, for instance, the gender gap behind different types of work (Table 3, column 2).

Youth who are engaged in paid work are 15.5% more likely to move across categories, girls’

odds are reduced by 15.8% (OR=0.842) if they are involved in paid work. Additionally, family

caring duties negatively amplify girls’ possibilities of transition across zones, from 7% to -

17.2%. Chores and family-related tasks are not detrimental for girls; however, being a girl and

Indigenous (based on the language spoken at home) is negatively associated with transition

(around 5%). Parental emphasis on youth’s education and schooling transmission mechanisms
17
from mothers to youth are weaker for girls than boys; nevertheless, literate mothers boost

positive repercussions on progression across zones of 8.7%, yet for the whole sample, this

estimate is much larger (= 61.5%).

A supplementary understanding of performance gaps can be obtained by estimating the

̂𝑚 ) of the dependent variable by profiles of

probability for each of the m categories (Pr

disadvantage (see Eq. (3)). These probabilities are included in Table 4. Firstly, results from

Panel A –showing profiles by youth’s gender plus education and Indigeneity (language)

variables– highlight either the higher chances of falling into zones 1 and 2 or the lesser chances

̂ for the group “Indigenous-mother illiterate-not

of reaching zones 4 and 5 when contrasting Pr

emphasis on education” versus the counterpart group. Positive gaps of 10%-12% (never been

to school)and 6%-7% (primary dropouts) occur between the two groups, and estimates indicate

negative gaps of 9% and 7% for primary leavers and secondary dropouts. All in all, language

spoken at home and mother’s education (proxied by literacy) are influential factors predicting

the zone of exclusion a youth (regardless of gender) will ultimately end up.

[Table 4 here]

Secondly, Panel B includes estimated probabilities based on combinations of work

conditions with gender. Here, we find analogous configurations for the gap between working

and non-working youth groups on the likelihood to fall into given zones of exclusions; though

smaller than in Panel A, gaps are more prominent for girls than for boys. For instance, girls are

7% more likely to have never enrolled into school if they are engaged in paid work, care work

or if they work for the family/farm than if they do not, whereas this gap is lower in the case of

boys (at 3%).

5.3 Research question 3 – impact of school, financial support and schooling conditions

We now move onto a different kind of research question, more directly aligned with the policy

domain. Regarding the youth group who ever attended some type of education, we investigate
18
what is the role of different conditions and various financial support types in enhancing the

chances of moving across zones. As mentioned above, our model is based on the ever attended

youth sample with discrete outcome taking four values, from zones 2, 3, 4 and a final zone

combining zones 5 and 6. Estimates15 for the MEs (based on Eq. (4)) are displayed in Table 5.

Panel A (Table 5) presents the MEs for school-related variables.16 The impact of language

across the four categories for the covariate “language first learnt to read” are more significant

than for “language of instruction”, thereby suggesting the narrower scope of language policies

to offset language (Indigeneity) as a marker of disadvantage once youth made into the

education systems. Learning to read in a local language, compared to learning to read in the

language of the test, is positively associated with primary dropout (by 6.2%) or being enrolled

at grade or below (by 2.3%), whereas the probabilities of falling into zone 4, 5 and 6 are reduced

by 4.4% and 4.1%, respectively. Being instructed in the language of the test, conversely, leads

to a shift of probabilities of 1.4% for the two latter categories, and adverse changes for zones

2 and 3 (of around 1%-2%), which are relatively small compared to the impact of the language

first learnt to read. Moreover, as expected, primary repetition is a strong precursor of dropout,

increasing dropout (column 1) and overaged/enrolled at earlier grades (column 2) by 10.2%

and 5.4%, though having reverse signs for the MEs for the top two zones. We do not find any

significant effects for truancy (missing school) and school location/travel time, which, although

significant, its impact across categories is moderately small (below 1%).

[Table 5 here]

In Panel B (Table 5), we present estimated MEs for various kind of support for youth that

is geared, in principle, toward leveraging education transition across zones. The overreaching

findings from these variables can be classified into two groups. Firstly, when support is

primarily meant to address youth’s income deprivation, estimates entail to some extent a lack

of connection of this support to schooling because most positive MEs (whether receiving

19
support or not) are for zones 2 or 3; this is the case of child/family grants, school meals or

vouchers for food (with MEs of around 0.005-0.027 and negative MEs for the other zones).

Secondly, when support is more closely linked to attendance or lowering its costs (i.e., cash

transfer to attend, scholarships, removal of fees), youth in school who had at least attended

above grade 5 (zone 4 and zones 5-6) are benefitted since MEs are positive and relatively large

across these two zones (e.g., for scholarship MEs are 0.043 and 0.064). Finally, amongst school

conditions under which youth may be able to progress and make it into the top exclusion zones

(Panel C, Table 5), estimates show that better quality of teaching, absence of discrimination

and lack of school fees (against the counterparts), would boost youth’s chances of concluding

primary education or beginning secondary by roughly 1.4% to 6.1%.

6. Conclusions

The out-of-school assessment of PISA for development (PISA-D) allowed us to gain insights

on the situation faced by OOSY from poor education systems with large populations of

marginalised students facing inequalities stemming from Indigeneity, rurality, poverty and

gender. In this paper, we employed the new OOSY survey (from PISA-D) comprising

students/youth aged 14-16 from five countries from the global south (i.e., Guatemala,

Honduras, Panama, Paraguay and Senegal) providing answers on the reasons behind access

and retention as well as differentiating between zones of exclusion. These zones are: zone 1

(never enrolled), zone 2 (primary dropout), zone 3 (enrolled but at grade 6 or below), zone 4

(leave at grade 6, primary leavers), zone 5 (dropout at secondary) and zone 6 (fading out,

enrolled but overaged). Specifically, we investigated which factors determine where youth falls

into the different zones and whether there is a differential impact of these factors by gender.

The leading findings of the paper are as follows.

Firstly, we found that drivers such as household wealth, parental education and

involvement in their children’s education can have wide-reaching impacts on progression

20
across zones. For example, youth from poor households are nearly 21% less likely to move up

across zones, whereas youth whose mothers are literate have 65% higher chances to move up

across zones. These impacts, however, are found to be of oppositive direction for bottom and

top zones of exclusion. For instance, parental education has increasing effect for zones 4 or 5

(between 3%-5%) but a negative impact (of 2%-7%) for those youth who only make it into

zones 1 or 2 (never attended or dropout in the first grades of primary). We also found that the

likelihood of moving across educational zones is around 5% higher for girls than boys.

Secondly, our results suggested that key markers of inequality have a gender dimension. This

holds when we scrutinised the impact that parental education, Indigeneity, and different types

of work have on girls versus boys' chances of falling into zones of exclusion. For instance, we

found that girls’ chances to move across categories of exclusion are reduced by around 16% if

they are engaged in paid work, while they are 7% more likely to have never enrolled in school

if also involved in other types of work (care or family work); for boys, the impact is much

lower (of 3%). Indigeneity for girls has relatively larger harmful effects on transition (of 5%)

and, too, mechanisms of educational transmission are weaker for girls in comparison to boys.

6.1. Policy implications

Indigeneity and language. The analysis showed that those learning in their native language are

further disadvantaged compared to those who learn in the test’s language. This highlights the

inexistence of a culturally-specific education and the need of bilingualism in delivery,

especially for those who do not speak the test language at home. This opens up a whole set of

possibilities for creating specific school material in Indigenous language as well as the training

and support of native teachers so that they are able to mediate Indigenous language education

with mainstream languages and cultures.

Financial aid. With regards to school conditions and policies, we found that, linking

financial support to attendance, boosting teaching quality as well as addressing discrimination,

21
have the potential to boost progression, extending the time youth may spend in education

systems. Concretely, estimates showed that financial support works more efficiently when

closely connected to attendance, through lowering or removing schooling costs, whereas

financial aid is less effective if applied to broad household deprivation (i.e., family grants). In

the former case, the impact of financial support led to increases of 4%-6% in finishing or

reaching secondary education; instead, when focusing on household’s deprivation, the effects

are largely found to be negative for top zones of exclusion. Also, raising teaching quality and

mitigating discrimination led to increasing a youths chance of reaching the end of primary or

beginning of secondary by 1.4% to 6.1%.

Family and education. Educated mothers enhance children’s possibilities to complete the

cycle of education, especially at secondary level. Women’s autonomy is positively related to

educational success. What might enhance girls’ future decision-making autonomy is the

insertion, at early stages, of holistic programs on sexuality, reproductive health, and physical

changes arising during adolescence. This could positively affect educational access and

retention.

Early warning systems. The model of six zones of exclusions needs to be continuously

updated and applied to specific contexts to target individuals at risk of dropout. Official data

(see Online Appendix E) are sometimes scarce and fail at reaching numerous fragile categories.

Monitoring students' progression within education systems and following youth’s life path is

fundamental in gathering information for successful interventions to promote educational

retention.

6.2. Limitations

The empirical analysis has some data limitations. Extrapolating policy implications from the

paper’s leading findings are made with caution, particularly given the cross-section nature of

the data (weaking causality claims as in other studies) and the presence of small sample sizes
22
which may affect predictions and the precision of MEs behind estimates for some sub-

categories. Linked to this is the lack of over-representation in the sample for those groups which

are more prone to be out of school (e.g., those living in poor rural areas and those living in

informal urban settlements facing constrained educational supply). Besides, we recognise the

fact the framing of youth progress based on the six zones of exclusion model is artificial, and

some groups within this model (i.e., repeaters and those not attending regularly) may need a

different conceptualisation and data.

23
 Conflict of interest

No potential conflict of interest was reported by the authors.

Acknowledgements

We would like to thank the OECD for offering the PISA-D (OOSY) dataset online at:

http://www.oecd.org/pisa/pisa-for-development/database/. We are grateful for the insightful

and critical comments from two referees which significantly improved the paper.

24
 References

Addey, C. (2020). The appeal of PISA for Development in Ecuador and Paraguay: theorising
and applying the global ritual of belonging. Compare: A Journal of Comparative and
International Education, 50(8), 1159-1174.
Addey, C. (2017). Golden relics & historical standards: How the OECD is expanding global
education governance through PISA for Development. Critical Studies in Education, 58(3),
311-325.
Addey, C., & Gorur, R. (2020). Translating PISA, translating the world. Comparative
Education, 56(4), 547-564.
Aldeman , M., Haimovich, F., Ham, A., & Vazquez, E. (2018). Predicting school dropout with
administrative data: new evidence from Guatemala and Honduras. Education Economics,
26(4), 356-372.
Aldeman, M., & Székely, M. (2017). An overview of school dropout in central America:
unresolved issues and new challenges for educational progress. European Journal of
Educational Research, 6(3), 235-259.
Ananga, E. (2011). Typology of school dropout: The dimensions and dynamics of dropout in
Ghana. International Journal of Educational Development, 31(374).
Arceo-Gómez, E., & Campos-Vázquez, R. (2014). Teenage pregnancy in Mexico: Evolution
and consequences. Latin American Journal of Economics, 51(1), 109-146.
Asil, M., & Brown, G. T. (2016). Comparing OECD PISA reading in English to other
languages: Identifying potential sources of non-invariance. International Journal of
Testing, 16(1), 71-93.
Auld, E., Xiaomin, L., & Morris, P. (2020). Piloting PISA for development to success: an
analysis of its findings, framework and recommendations. Compare: A Journal of
Comparative and International Education, 1-25.
Auld, E., Rappleye, J., & Morris, P. (2019). PISA for Development: how the OECD and World
Bank shaped education governance post-2015. Comparative Education, 55(2), 197-219.
Azevedo, J. P., Hasan, A., Goldemberg, D., Iqbal, S. A., & Geven, K. (2020). Simulating the
potential impacts of COVID-19 school closures on schooling and learning outcomes: A set
of global estimates. Policy research working paper 9284. The World Bank.
Barber, S., Diex Roux, A. V., Cardoso, L., Santos, S., Toste, V., James, S., . . . Chor, D. (2018).
At the intersection of place, race, and health in Brazil: Residential segregation and cardio-
metabolic risk factors in the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil).
Social Science & Medicine, 199, 67-76.
Batz, G. (2018). The Ixil University and the decolonisation of knowledge. In L. Smith, E. Tuck,
& K. Yang (Eds.), Indigenous and Decolonizing Studies in Education: Mapping the Long
View (pp. 103–115). Routledge.
Behrman, J., de Hoyos, R., & Székely, M. (2014). Conceptual framework for approaching the
“Out of School and Out of Work” youth phenomenon”. Mimeo, Washington DC.
Bowers, A., & Sprott, R. (2012). Why tenth graders fail to finish high school: A dropout
typology latent class analysis. The Journal of Education for Students Placed Atomic Risk
(JESPAR), 17(3), 129–148.

25
Bray, M., Kobakhidze, M. N., & Suter, L. E. (2020). The challenges of measuring outside-
school-time educational activities: Experiences and lessons from the Programme for
International Student Assessment (PISA). Comparative Education Review, 64(1), 87-106.
Camara, A., & Adam, S. (2012). Challenges to primary school enrolment in the Gambia.
Journal of Social, Political and Economic Studies, 37(3), 347–366.
Carr-Hill, R. (2015). PISA for Development Technical Strand C: Incorporating Out-of-School
15-Year-olds in the Assessment”, OECD Education Working Papers, No. 120, OECD
Publishing, Paris
Carr-Hill, R. (2012). Finding then counting out of school children. Compare: A Journal of
comparative and international education, 42(2), 187-212.
Chaaban, J., & Cunningham, W. (2011). Measuring the economic gain of investing in girls:
The girl effect dividend. World Bank Policy Research Working Paper Series, Policy
Research Working Paper.
Daly, A., Mbenga, B., & Camara, A. (2016). Barriers to participation and retention: engaging
and returning 'out of school' children in The Gambia. Education 3-13, 44(2), 181-196.
Davis, J. (2016). ¿Educación o desintegración? Parental migration, remittances and left-behind
children’s education in Western Guatemala. Journal of Latin American Studies, 48(3), 565–
590.
Delprato, M., Akyeampong, K., & Dunne, M. (2017). Intergenerational education effects of
early marriage in Sub-Saharan Africa. World Development, 91, 173-192.
Delprato, M., & Akyeampong, K. (2017). The effect of early marriage timing on women's and
children's health in Sub-Saharan Africa and Southwest Asia. Annals of global health, 83(3-
4), 557-567.
Delprato, M., Akyeampong, K., Sabates, R., & Hernandez-Fernandez, J. (2015). On the impact
of early marriage on schooling outcomes in Sub-Saharan Africa and South West
Asia. International Journal of Educational Development, 44, 42-55.
de Sousa Santos, B. (2012). Cuando los excluidos tienen Derecho: justicia indígena,
plurinacionalidad e interculturalidad. Justicia indígena, plurinacionalidad e interculturalidad
en Ecuador, 13-50.
Dixon-Mueller, R. (2008). How young is ‘too young’? Comparative perspectives on adolescent
sexual, marital, and reproductive transitions. Studies in Family Planning, 39(4), 247–262.
Fischman, G. E., Topper, A. M., Silova, I., Goebel, J., & Holloway, J. L. (2019). Examining the
influence of international large-scale assessments on national education policies. Journal of
education policy, 34(4), 470-499.
Fortin, L., Marcotte, D., Potvin, P., Royer, E., & Joly, J. (2006). Typology of students at risk of
dropping out of school: Description by personal, family and school factors. European
Journal of Psychology of Education, 21(363).
Frazelle, S., & Nagel, A. (2015). A Practitioner’s Guide to Implementing Early Warning
Systems.
Grasso, K. (2018). Guatemalan youth and education: family, environment, and dropping out.
Latin American, Caribbean, and U.S. Latino Studies Honors Program, 1.

26
Guinan, J. (2015). Nearly 20 years after peace pact, Guatemala’s women relive violence. CNN
Report of April 7 available at: https://www.cnn.com/2015/04/02/world/iywguatemala-
gender-violence/index.html.
Hajny, K. (2015). Water and wealth: A Guatemalan case study. Journal of Economics and
Economic Education Research, 16(2), 119-136.
Hannum, E., & Adams, J. (2007). Girls in Gansu, China: Expectations and aspirations for
secondary schooling. In M. Lewis, & M. Lockheed (Eds.), Education, gender and
education: Case studies from the developing world (pp. 71-98). Washington, DC: Center for
Global Development.
Hendrick, E. C., & Marteleto, L. (2017). Maternal household decision-making autonomy and
adolescent education in Honduras. Popul Res Policy Rev, 36, 415-439.
Johnston, D., Schurer, S., & Shields, M. (2014). Maternal gender role attitudes, human capital
investment, and labour supply of sons and daughters. Oxford Economic Papers, 66(3), 631-
659.
Kabeer, N., & Santos, R. (2017). Intersecting Inequalities and the sustainable development
goals: insights from Brazil. Working paper, 14.
Kameyama, Y. (2018). Out-of-school children and disability. JICA Research Institute, 1-32.
Lehman, D. (2003). Bringing the school to the children: shortening the Path to EFA.
Washington, DC: World Bank.
Lewin, K. M. (2007). Improving access, equity and transitions in education: Creating a research
agenda. University of Sussex: Centre for International Education.
Lewin, K. M. (2009). Access to education in sub‐Saharan Africa: patterns, problems and
possibilities. Comparative Education, 45(2), 151-174.
Lewin, K. (2011). Making rights realities: Researching educational access, transitions and
equity. CREATE Synthetic Report. Centre for International Education, University of
Sussex.
Lewin, K. M.(2013). Does privatising educational services for the poor make sense? Paper
presented at the conference on the impact of privatisation and marketisation on the
education of disadvantaged children, (pp. 1-18).
Lewin, K. M. (2017). Making rights realities. Privatisation and the Education of Marginalised
Children: Policies, Impacts and Global Lessons, 81.
Lloyd, C., & Mensch, B. (2008). Marriage and childbirth as factors in dropping out from
school: An analysis of DHS data from Sub-Saharan Africa. Population Studies, 62(1), 1-13.
Long, J. S., & Freese, J. (2014). Regression models for categorical dependent variables using
Stata. Third edition, Stata Press.
Menjívar, C. (2008). Violence and women's lives in Eastern Guatemala: A conceptual
framework. Latin American Research Review, 43(3), 109-136.
Meyer, H. D., & Benavot, A. (Eds.). (2013). PISA, Power, and Policy: The Emergence of
Global Educational Governance. Oxford: Symposium.
Ministry of Basic and Secondary Education (MoBSE). (2009). The special needs education and
inclusive policy framework 2009–2015. Banjul: Republic of the Gambia.

27
Mulkeen, A., & Chen, D. (Eds.). (2008). Teachers for rural schools: experiences in Lesotho,
Malawi.
Murphy-Graham, E., Cohen, A., & Pacheco-Montoya, D. (2020). School dropout, child
marriage, and early pregnancy among adolescent girls in rural Honduras. Comparative
Education Review, 64(4), 703-724.
Nichols, B. (2021). Nothing is easy: educational striving and migration deferral in Guatemala.
Journal of Ethnic and Migration Studies, 2-17.
OECD (2018a). PISA for development assessment and analytical framework: Reading,
mathematics and science.
OECD (2018b). PISA for development. Technical report.
Ogrodnik, C., & Borzutzky, S. (2011). Women under attack: Violence and poverty in
Guatemala. Journal of International Women's Studies, 12(1), 109-146.
Orenstein, H. (2017). These portraits shed light on the teen pregnancy epidemic in Latin
America. Retrieved from https://blog.malala.org/these-stunning-portraits-shed-light-on-the-
teen-pregnancy-epidemic-in-latin-america-3146196976a1
Orozco, M., & Valdivia, M. (2017). Educational challenges in Honduras and consequences for
human capital and development. Inter-American Dialogue.
Peters, S. (2010). Review of marginalisation of people with disabilities in Lebanon, Syria and
Jordan. Background paper prepared for the Education for All Global Monitoring Report
2010: reaching the marginalised.
Pôle de Dakar, & UNESCO BREDA. (2005). EFA in Africa: Paving the Way for Action.
Senegal: Pôle de Dakar and UNESCO BREDA.
Richardson, E., Phillips, M., Colom, A., Khalil, I., & Nichols, J. (2019). Out of school factors
affecting indigenous girls' educational attainment: a theory of change for the opening
opportunities program in rural Guatemala. Comparative and International Education, 47(2),
1-17.
Seeley, C. (2018). The pursuit of education by women in rural Honduras. Journal of Research
Initiatives, 4(2), 1-18.
Silver, H. (2007). Social exclusion: comparative analysis of Europe and Middle East youth. The
Middle East Youth Initiative: Working paper 1.
Singh, R., & Mukherjee, P. (2018). Push Out, Pull Out, or Opting Out? Reasons Cited by
Adolescents for Discontinuing Education in Four Low- and Middle-Income Countries. In J.
Lansford, & P. Banati (Eds.), Handbook of adolescent development research and Its impact
on global policy. Oxford: Oxford University Press.
Tarallo, L. (2019). Gender inequality in Guatemala: why girls receive less education than boys.
Global Majority E-Journal, 10(1), 43-53.
Tröhler, D. (2013). The OECD and Cold War Culture: Thinking Historically about PISA. In
PISA, Power and Policy: The Emergence of Global Educational Governance, ed. H. D.
Meyer and A. Benavot. Oxford: Symposium.
Tumen, S. (2015). Skill acquisition in the informal economy and schooling decisions: Evidence
from emerging economies. Labour, 29(3), 270-290.
UIS (2019). New Methodology Shows that 258 Million Children, Adolescents and Youth Are
Out of School. Fact Sheet no. 56
28
UNESCO. (2014). Out of school children. Retrieved May 13, 2020, from
http://uis.unesco.org/apps/visualisations/oosci-data-tool/index-en.html#en/SEN
UNESCO (2020a). Global education monitoring report 2020: inclusion and education: all
means all.
UNESCO (2020b). A new generation: 25 years of efforts for gender equality in education.
Gender Report of the Global Education Monitoring Report series.
UNICEF. (2015). Factores Asociados al Abandono del Sistema Educativo en la Transición
Escolar. Panama.
UNICEF (2018). The out-of-school children Initiative. Formative evaluation of the out of
school children initiative (OOSCI). New York: UNICEF. Retrieved from link
USAID. (2019). The future of the Honduran education system: a macro analysis of education
quality and quantity's impact on development.
Vayachuta, P., Ratana-Ubol, A., & Soopanyo, W. (2015). The study of 'out-of-school' children
and youth situations for developing a lifelong education model for 'out-of-school' children
and youth. SHS Web of Conferences, 1-5.
Xiaomin, L. (2021). PISA for Development: Redefining Out-of-school Youth for Success.
NORRAG Highlights blog. Retrieved from link
Xiaomin, L., & Auld, E. (2020). A historical perspective on the OECD’s ‘humanitarian turn’:
PISA for Development and the Learning Framework 2030. Comparative Education, 56(4),
503-521.
Ward, M. (2020). PISA for Development: Out-of-school assessment: Results in Focus.
Ward, M. (2018.) PISA for Development: Results in focus.
Ward, M., & P. Zoido (2015). PISA for Development. ZEP 4 (15): 21–25.
Willms, J. D. (2018). Learning Divides: Using Monitoring Data to Inform Education
Policy. Montreal: UNESCO Institute for Statistics.
World Bank. (2020). The COVID-19 pandemic: Shocks to education and policy responses.

29
Additional Information – Online Appendix

Additional information can be found in the online appendix of this article:

Diagram A1. Five Dimensions of Exclusion (5DE)

Table A1. Outcomes and covariates description (main specification)
Table B1. Zones of exclusion rates per country
Figure B1. Predicted probabilities for six zones of exclusion
Table C1. Estimated probabilities by the dependent variable values (six zones of exclusion)
per country and by gender
Table D1. Ordered logit estimates (OR) for “ever attended sample” (outcome: four zones of
exclusion)
Appendix E – Estimations, official data and households surveys

30
Figure 1. Ordered logit (six zones of exclusion: Z1,…,Z6) marginal effects for selected covariates
(full model)

31
Table 1. Descriptive statistics for outcomes (zones of exclusion) and covariates used in the analysis
Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6
Never Primary Enrolled at Primary Secondary
enrolled dropouts grade 6 or below leavers dropouts Fading out
(1) (2) (3) (4) (5) (6)
Panel A - outcomes
Rate 13% 15% 22% 24% 21% 4%
N 1,119 1,114 2,323 1,605 1,704 206

Panel B - covariates
Students/Youth
age 14.95 15.00 14.56 14.94 15.12 15.19
gender - female 0.46 0.50 0.45 0.53 0.49 0.44
disability 0.04 0.07 0.07 0.04 0.04 0.05
language of the test (no) 0.39 0.40 0.49 0.33 0.32 0.48
work - paid 0.18 0.33 0.00 0.30 0.21 0.37
work - care 0.46 0.45 0.49 0.37 0.38 0.46
work- chores 0.84 0.89 0.89 0.88 0.83 0.84
work - family, farm 0.39 0.30 0.35 0.24 0.20 0.27
well-being - index 0.099 -0.191 -0.044 0.100 -0.004 -0.225
health - index -0.051 0.067 0.120 -0.072 -0.051 0.122
Family
poor 0.57 0.66 0.52 0.61 0.29 0.32
live with parents 0.66 0.67 0.66 0.74 0.63 0.53
mother - literate 0.20 0.26 0.25 0.37 0.60 0.59
father - literate 0.33 0.38 0.39 0.49 0.69 0.69
finish education - important 0.42 0.52 0.71 0.62 0.76 0.81
safety - index -0.398 -0.165 -0.079 0.037 0.093 0.074

Notes: (1) Weighted average means of outcomes and covariates across different zones of exclusion. (2) For a working definition of covariates, see Online Appendix
(Appendix A, Table A1).

32
Table 2. Ordered logit estimates (outcome: six zones of exclusion)
OR % %StdX % %StdX
(1) (2) (3) (4) (5) (6)
age 1.118*** 1.169*** 11.8 9.5 16.9 13.6
gender - female 1.019*** 1.053*** 1.9 0.9 5.3 2.6
disability 1.094*** 1.001 9.4 2.1 0.1 0
language of the test (no) 1.193*** 1.411*** 19.3 9.2 41.1 18.8
work - paid 1.016*** 1.077*** 1.6 0.7 7.7 3.1
work - care 0.838*** 0.851*** -16.2 -8.4 -14.9 -7.8
work- chores 1.425*** 1.326*** 42.5 11.1 32.6 8.7
work - family, farm 0.801*** 0.785*** -19.9 -10 -21.5 -10.9
well-being - index 1.051*** 1.047*** 5.1 4.6 4.7 4.2
health - index 1.115*** 1.078*** 11.5 9.5 7.8 6.5
family - poor 0.795*** -20.5 -10.5
live with parents 1.031*** 3.1 1.4
mother - literate 1.674*** 67.4 22.9
father - literate 1.455*** 45.5 19.9
finish education - important 2.024*** 102.4 41.8
safety - index 1.164*** 16.4 14
Honduras 1.598*** 1.793*** 59.8 15.4 79.3 19.6
Panama 3.608*** 2.345*** 260.8 16.2 134.5 10.5
Paraguay 3.087*** 2.160*** 208.7 29.8 116 19.5
Senegal 0.417*** 0.556*** -58.3 -34.9 -44.4 -25
𝜏1 1.080*** 3.671***
𝜏2 2.951*** 10.716***
𝜏3 10.120*** 39.103***
𝜏4 43.013*** 175.041***
𝜏5 167.166*** 709.576***
N 7,887 7,887

Notes: (1) Weighted ordered logit regression. (2) Countries base category: Guatemala. (3) % is equal to OR-1,
and %StdX is the proportional change on OR by a standard deviation increase on X (continuous).
Significance levels: * p < 0.10, **p < 0.05, ***p < 0.01

33
Table 3. Ordered logit estimates (outcome: six zones of exclusion) - gender interacted with language,
work and education variables
OR % %StdX
(1) (2) (3)
age 1.167*** 16.7 13.4
disability 1.003 0.3 0.1
well-being - index 1.046*** 4.6 4.1
health - index 1.078*** 7.8 6.5
gender - female 0.918*** -8.2 -4.2
language of the test (no) 1.466*** 46.6 21
female x language of the test (no) 0.946*** -5.4 -2.3
work - paid 1.155*** 15.5 6.1
female x work - paid 0.842*** -15.8 -4.4
work - care 0.930*** -7 -3.5
female x work - care 0.828*** -17.2 -8.3
work - chores 1.225*** 22.5 6.2
female x work - chores 1.295*** 29.5 13.8
work - business, farm 0.761*** -23.9 -12.1
female x work - business, farm 1.071*** 7.1 2.4
mother - literate 1.615*** 61.5 21.1
female x mother - literate 1.087*** 8.7 2.4
finish education - important 2.004*** 100.4 41.1
female x finish education - important 1.019*** 1.9 0.9
family - poor 0.797*** -20.3 -10.4
live with parents 1.033*** 3.3 1.5
father - literate 1.454*** 45.4 19.9
safety - index 1.163*** 16.3 13.9
N 7,887

Notes: (1) See notes in Table 2. (2) Specification also includes country dummies as in Table 2.
Significance levels: * p < 0.10, **p < 0.05, ***p < 0.01

34
Table 4. Estimated probabilities by the dependent variable values (six zones of exclusion) based on gender's interactions
Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6
Never Primary Enrolled at grade Primary Secondary Fading
enrolled dropouts 6 or below leavers dropouts out
(1) (2) (3) (4) (5) (6)
Panel A
female, Indigeneity, mother literate, finish
education - important
yes, yes, no, no 0.217 0.21 0.281 0.201 0.067 0.025
yes, no, yes, yes 0.101 0.136 0.268 0.296 0.138 0.061
no, yes, no, no 0.215 0.209 0.281 0.202 0.067 0.025
no, no, yes, yes 0.114 0.148 0.275 0.284 0.124 0.054

Panel B
female, work paid, work care, work chores, work
family/farm
yes, yes, yes, yes, yes 0.211 0.199 0.271 0.209 0.077 0.031
yes, no, no, yes, no 0.145 0.165 0.272 0.257 0.111 0.05
no, yes, yes, yes, yes 0.192 0.192 0.276 0.222 0.084 0.034
no, no, no, yes, no 0.165 0.178 0.276 0.242 0.097 0.041

Notes: (1) Estimated probabilities are estimated using the same specification as in Table 3, and obtained by using the Stata command mtable (Long and Freese, 2014).

35
Table 5. Ordered logit estimates marginal effects (MEs) for ever attended sample (outcome: 4 zones
of exclusion)
Zone 2 Zone 3 Zone 4 Zone 5-6
Enrolled at Secondary
Primary grade 6 or Primary dropouts,
dropouts below leavers Fading out
(1) (2) (3) (4)
Panel A - school variables
not language of the test - learnt to read 0.062 0.023 -0.044 -0.041
instruction - language of the test -0.019 -0.009 0.014 0.014
missed school (more than 3 months) (+ns) -0.001 0.000 0.001 0.001
primary repetition 0.102 0.054 -0.072 -0.084
travel time to school - longer -0.008 -0.004 0.006 0.006

Panel B - support for attendance

child or family grants to attend 0.024 0.010 -0.017 -0.018
cash transfer to attend school (+) -0.003 -0.001 0.002 0.002
scholarship -0.064 -0.043 0.043 0.064
school meals 0.027 0.013 -0.019 -0.021
vouchers or coupons for food 0.011 0.005 -0.007 -0.008
waiver of school fees -0.005 -0.002 0.003 0.004
school books or notebooks 0.004 0.002 -0.003 -0.003

Panel C - conditions for continuing school

continue school - if closer to home 0.012 0.006 -0.008 -0.009
continue school - if better quality of teaching -0.091 -0.028 0.061 0.058
continue school - if safer school (+ns) 0.002 0.001 -0.001 -0.001
continue school - if no discrimination -0.019 -0.008 0.013 0.014
continue school - if accepting students
differences 0.012 0.006 -0.008 -0.010
continue school - if financial incentives (+ns) 0.000 0.000 0.000 0.000
continue school - if no fees -0.009 -0.004 0.006 0.006

Notes: (1) All marginal effects (from 0 to 1) are statistically significant at 1%, expect from (+ns) denoting non-
significant effects and (+) significant at 10%. (2) Each specification obtains binary marginal effects using as set
of covariates those included in each Panel plus country dummies (using sampling weights). (3) Command used
to calculate MEs: mchange (Long and Freese, 2014).

36
Notes

1
We have capitalised the word “Indigeneity" following common practice in the field, which aims to highlight
that the concept of Indigeneity is unique, encompassing cultural aspects needed to be considered for the political
international visibility of this group.
2
For instance, Tumen (2015) reports how informal jobs are a high deterrent to educational retention among
emerging economies.
3
For examples on multicultural and pluractional experiences addressing Indigenous discrimination, see: de
Sousa Santos (2012).
4
This dataset has been recently released online by the OECD (December 2020) and it can be found at:
https://www.oecd.org/pisa/pisa-for-development/database/.
5
For a description and critic of the OOSY component of PISA-D, see the recent NORRAG blog at:
https://www.norrag.org/pisa-for-development-redefining-out-of-school-youth-for-success-by-li-xiaomin/. The
strands are: (a) Strand A – learning test items targeted at the lower end of the performance distribution; (b)
Strand B – contextual questionnaires to capture the different barriers faced by students in low- and middle-
income countries ; (c) Strand C – an approach to incorporate the OOSY into the assessment. Two immediate
objectives from PISA-D are capacity building for management and use of the corresponding datasets and the use
of PISA-D for monitoring and achievement of the Sustainable Development Goal for education (SDG 4).
6
A further language concern is whether PISA-D (and the OOSY) gives room for children to answer main
learning questions on their own mother tongue; this is ambiguous. And, indeed, this coincides with the absence
of policy implications related to the learning gap between mother tongue speakers and native speakers across the
array of PISA-D reports (Auld, et al. 2020) and issues related to rights for disadvantaged speakers (most
Indigenous groups already historically segregated in the LAC region for example) and the quality of data
robustness by ignoring mother tongue testing.
7
The blog can be found at: https://www.norrag.org/pisa-for-development-redefining-out-of-school-youth-for-
success-by-li-xiaomin/
8
We exclude 583 observations due to questionnaires being uncompleted. We exclude 583 observations due to
questionnaires being uncompleted. Note that final sample of around 7,900 observations is lower than the
corresponding in-school sample from PISA-D (around 19,500 for all countries excluding Panama). Yet the
OOSY sample size is reasonable considering the out-of-school rates for adolescents for these countries (ranging
between 3% in Paraguay to 27% in Guatemala and Honduras and 39% in Senegal). See WIDE database.

9
See Diagram A1 (Online Appendix) for details.
10
Table B1 (Online Appendix) shows that there is an important between-country variation. In particular, lack of
access and primary overage are the main problem in Senegal (of 29.7% and 41%, respectively), while in the
other four LAC countries there are major problems in primary dropouts or leavers (Guatemala and Honduras
between 21%-42%) and relatively wealthier countries (Panama and Paraguay) have education flows bottlenecks
at secondary (dropouts rates of 31%-42%).
11
When assessing youth’s progression gaps based on language of instruction versus language of the test, we are
aware of the limitation behind this exercise which is far more complex (e.g., several languages recognised at
school and spoken in similar communities). Language is indeed a locus of struggle, where multiple identities
intersect and would need further examination. Though a finer examination of this issue is beyond the scope of
the paper due to its complexity and, at the same time, the main objective of the paper being a comparative effort.
Still, the paper analyses multiple educational barriers, and the criterion adopted "language of instruction vs
language of the test" already encompasses a high percentage of the struggle students might face. We thank an
anonymous referee for pointing this out.
12
The distribution of the estimated probabilities for the six zones of exclusions are plotted in Figure B1 (Online
Appendix). This figure shows that most predictions for the two extremes categories (zone 1 and zone 6) tend to
37
be less than 0.10, with most predictions for the middle categories (zones 3 and zone 4) being above 0.25, and
with zone 2 having the largest dispersion for its estimated probabilities (between 0.05-0.25).
13
The likelihood-ratio test leads to a rejection of the null hypothesis (simpler model embedded within fuller
model) because its 𝜒62 = 109697.61 (p-value = 0.00). Additional tests such as the parallel regression assumption
was accepted, and other fitness statistics were calculated (AIC, BIC and different types of 𝑅2 ) which are
available from the authors upon request.
14
Estimated probabilities for the six zones across the five countries by gender can be found in Table C1 (Online
Appendix).
15
Note that, across all categories, the sum of average MEs are equal to zero because an increase in the
probability of on category should be counterbalanced by a decrease in another category.
16
For the full specification estimates (OR), see Table D1 (Online Appendix).

38
 ONLINE APPENDIX
Zones of Educational Exclusion of Out-of-School Youth

1
APPENDIX A

Diagram A1. Five Dimensions of Exclusion (5DE)

Source: UNICEF (2018)

2
Table A1. Outcomes and covariates description (main specification)
Variables Description
Outcomes
zones of exclusion 1 - never attended school; 2 - primary dropouts (drop from primary school between grades 1 to 5);
3 - currently enrolled but a grade according to student's age (grade 6 or below); 4 - primary leavers
(dropout in grade 6); 5 - dropout at secondary level; 6 - fading out (attending grade 7 or above)
Covariates
1) Student/Youth
age Youth's age
gender =1 if female, =0 if male
disability =1 if has disability, 0 otherwise
language spoken most of the time =0 if speak language of test, =1 otherwise
work - paid Work and paid (at least 10 hours), 1=yes, 0=no
work - care Work: care for household members (last week), 1=yes, 0=no
work- chores Work: different chores (last week), 1=yes, 0=no
work - family, farm Work: farm or family business (last week), 1=yes, 0=no
well-being - index Variables included for the construction of the index (by factor analysis) coded as 1 = never or almost never,
2 = about once a week, 3 = 2 0r 3 times a week, 4 = almost every day
a) You are too fearful or nervous
b) You are afraid that other people will think you are stupid
c) You worry more than most people
d) You worry about what other people think of you
e) You cry without a good reason
f) You feel lonely
g) Other people seem to have more fun than you
h) You feel sad or depressed
i)You have trouble falling asleep at night
j) A lot of things seem to bother you

Continued next page

3
health - index As above for the index construction (factor analysis) and coding of variable. Variables included in index:
a) A chronic health problem
b) An infectious health problem
c) Gastrointestinal problems
d) A cold or flu
e) An injury that needed treatment
f) Pain that was long-lasting or recurring
g) Insomnia
h) Fatigue that was long-lasting or recurring
i) AIDS/HIV
2) Family
poor =1 if family ESCS index falls into bottom half of distribution, 0 otherwise
live with parents =1 if youth live with either mother or father, 0 otherwise
mother - literate =1 if mother is literate, 0 if she is non-literate or semi-literate
father - literate =1 if father is literate, 0 if he is non-literate or semi-literate
finish education - important Parents talk to you about the importance of completing secondary school, 0=never or hardly ever, 1 otherwise
safety - index School safety around home/school. Factors analysis based on the variables (1=very unsafe…, 4= very safe)
a) You felt safe at your school
b) You felt safe on your way to school
c) You felt safe on your way home from school

4
APPENDIX B

Table B1. Zones of exclusion rates per country

Guatemala Honduras Panama Paraguay Senegal
(1) (2) (3) (4) (5)
Zone 1 - Never enrolled
% 12.8 5.5 13.9 2.5 29.7
N 142 61 247 18 651
Zone 2 - Primary dropouts
% 24.1 20.6 5.4 14.1 13.8
N 323 229 141 114 307
Zone 3 - Enrolled at grade 6 or
below
% 18.3 17.0 13.9 20.0 41.0
N 705 193 238 360 827
Zone 4 - Primary leavers
% 31.0 42.3 17.2 22.1 8.2
N 379 518 368 186 154
Zone 5 - Secondary dropouts
% 9.0 12.1 49.2 31.2 4.7
N 120 230 1036 239 79
Zone 6 - Fading out
% 4.8 2.6 0.4 10.1 2.6
N 58 39 21 61 27

Figure B1. Predicted probabilities for six zones of exclusion

5
APPENDIX C

Table C1. Estimated probabilities by the dependent variable values (six zones of exclusion) per
country and by gender
Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6
Enrolled at
Never Primary grade 6 or Primary Secondary Fading
enrolled dropouts below leavers dropouts out
(1) (2) (3) (4) (5) (6)

Guatemala 0.137 0.179 0.312 0.255 0.085 0.031

female 0.134 0.177 0.312 0.258 0.087 0.032
male 0.139 0.182 0.312 0.252 0.083 0.031

Honduras 0.081 0.123 0.28 0.324 0.137 0.055

female 0.079 0.121 0.278 0.326 0.139 0.057
male 0.082 0.125 0.282 0.322 0.135 0.054

Panama 0.063 0.101 0.253 0.346 0.167 0.072

female 0.061 0.099 0.251 0.347 0.169 0.073
male 0.064 0.102 0.255 0.344 0.164 0.07

Paraguay 0.069 0.109 0.264 0.338 0.155 0.065

female 0.068 0.107 0.262 0.34 0.157 0.066
male 0.071 0.111 0.266 0.337 0.152 0.063

Senegal 0.221 0.232 0.299 0.18 0.051 0.018

female 0.217 0.23 0.3 0.183 0.052 0.018
male 0.225 0.234 0.297 0.177 0.05 0.017

6
APPENDIX D

Table D1. Ordered logit estimates (OR) for ever attended sample (outcome: four zones of exclusion)

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

age 1.356*** 1.365*** 1.371***
gender - female 1.126*** 1.160*** 1.151***
disability 1.069*** 1.061*** 1.038***
language of the test (no) 1.030*** 1.006 1.010**
work - paid 1.285*** 1.307*** 1.352***
work - care 0.835*** 0.828*** 0.810***
work- chores 1.257*** 1.248*** 1.198***
work - family, farm 0.991** 0.974*** 0.963***
well-being - index 1.140*** 1.155*** 1.149***
health - index 0.993*** 0.990*** 0.976***
family - poor 0.825*** 0.839*** 0.838***
live with parents 0.956*** 0.985*** 0.987***
mother - literate 1.417*** 1.456*** 1.466***
father - literate 1.516*** 1.565*** 1.552***
finish education - important 1.460*** 1.460*** 1.352***
safety - index 1.026*** 1.031*** 1.031***
not language of the test - learnt to read 0.706*** 0.748***
instruction - language of the test 1.118*** 0.967***
missed school (more than 3 months) 1.004 0.971***
primary repetition 0.530*** 0.544***
travel time to school 1.049*** 1.076***
child or family grants to attend 0.869*** 0.893***
cash transfer to attend school 1.015* 1.048***
scholarship 1.532*** 1.590***
school meals 0.851*** 0.868***
vouchers or coupons for food 0.939*** 0.931***
waiver of school fees 1.029*** 0.972***
school books or notebooks 0.976*** 0.945***
continue school - if closer to home 0.933*** 0.963***
continue school - if better quality of
teaching 1.645*** 1.522***
continue school - if safer school 0.990 1.047***
continue school - if no discrimination 1.118*** 1.077***
continue school - if accepting students
differences 0.930*** 0.901***
continue school - if financial
incentives 1.002 1.067***
continue school - if no fees 1.051*** 1.008
N 6,951 6,951 6,951 6,951 6,951 6,951

Notes: (1) Weighted ordered logit regression.

Significance levels: * p < 0.10, **p < 0.05, ***p < 0.01

7
 APPENDIX E – Estimations, official data and households surveys

The numbers regarding out-of-school youth (OOSY) are striking and worrying. However,

numerous publications show that the actual number of OOSY is underestimated, especially in

developing countries (Buettner, 2008). The latter is due to poor official and administrative data,

to the structural cadence of national censuses, unable to track the individual’s life trajectories

effectively, and, most alarmingly, due to the very status of many children. Carr-Hill (2012)

highlights how households surveys, one the most employed statistical method to gather

information about the broad population by international agencies, fails at gathering “data about

the poorest of the poor” (p.190). Such disadvantaged share of the population includes homeless

children, mobile, nomadic or pastoralist populations, displaced children in refugee camps,

slums inhabitants, children located in dangerous areas (Carr-Hill, 2012). One of the biggest

challenges to developing targeted and effective interventions, which would favour the efficient

allocation of funds, is, therefore, the development of up-to-date and comprehensive official

population data in order, for example, to target those categories most vulnerable to dropout

(Aldeman & Székely, 2017).

References

Aldeman, M., & Székely, M. (2017). An overview of school dropout in central America:
unresolved issues and new challanges for for educational progress. European Journal
of Educational Research, 6(3), 235-259.
Buettner, T. (2008). Preparing population estimates for all countries of the world:
Experiences and challenges. Paper presented at the Conference on Data Quality for
International Organisations, Committee for the Coordination of Statistical Activities,
Food and Agricultural Organisation.
Carr-Hill, R. (2012). Finding and then counting out-of-school children. Compare: A Journal
of Comparative and International Education, 42(2), 187-212.

View publication stats