To What Extent has Bilingual Education Increased Education Outcomes in Bolivia?

(1)

To What Extent has Bilingual

Education Increased Education

Outcomes in Bolivia?

Víctor López Hernández

Student Number: 10846506

August 2018

Supervisor: Dr. Hessel Oosterbeek

Track: Development Economics

Master Thesis

(2)

1 Statement of Originality

This document is written by Víctor López Hernández, who declares to take full responsibility

for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources

other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion

of the work, not for the contents.

(3)

2 To What Extent has Bilingual Education Increased Education Outcomes in

Bolivia?

Abstract

This investigation examines the impact of bilingual education in Bolivia, which has increasingly been used across developing countries as an alternative to the largely ineffective education interventions previously designed and deployed. The investigation examines the extent to which ASEP, a large-scale bilingual education program implemented in 2006, increased education outcomes for the population, females, rural and indigenous individuals. It exploits the fact that some areas used an indigenous language for instruction, while some maintained Spanish, to carry out a difference-in-difference analysis for the treatment and control groups before and after the reform. The results reveal that bilingual education was largely successful at increasing education outcomes: the control group had years of schooling, literacy, numeracy and primary schooling increase by 0.514 years, 8.6, 8.7 and 10.9 percent, respectively. Females, rural and indigenous individuals, who have traditionally been marginalized groups, also saw big improvements in their educational outcomes. These results are mostly significant at the 1 percent level and robust to falsification tests. As such, the results corroborate the pedagogical literature and empirical evidence on bilingual education, which claim that it increases teaching efficiency and demand for education, hence improving outcomes. This suggests that large-scale bilingual programs are also successful at increasing schooling and learning. Hence, future research could look at the potential efficiency gains from large-scale bilingual programs, which has often been overlooked, while policy makers in developing countries ought to expand bilingual education programs.

Víctor López Hernández

Development Economics University of Amsterdam

10846506

(4)

3

4

I. Introduction

Despite great expansion in recent years, education in developing countries remains limited and inadequate. Today, 263 million children remain out of school (UNESCO, 2018), many leave school at an early age and the deficient resources across schools and appalling teacher quality are some of the reasons for why children in developing countries remain illiterate and innumerate (World Bank, 2017). This dismal situation has inspired a range of interventions aimed at increasing education, typically through the expansion of access and availability of schooling and through the enhancement of school quality (Prevatt & Kelly, 2003). However, school quality has generally not improved and expanding school access has not generated any significant improvements in learning (Hanushek, 1995).

The inability of academics and policy makers to achieve significant improvements for children's education has led to a reshuffling in the design and implementation of education programs. For instance, scholars have increasingly looked at language of instruction to improve the quality and reach of education. In particular, language of instruction has been labeled as the key to improvements in psychosocial factors, enhanced teaching efficiency and increased community involvement. In developing countries where a large proportion of the population often speaks a language other than their ex-colonizers, this has inspired the take up of bilingual programs. A prime example can be found in Bolivia, which in 2006 passed the Avelino Siñani - Elizardo Pérez (ASEP) law, implementing bilingual education in schools across the country.

The evidence on bilingual programs in developing country settings would largely support the implementation of the ASEP reform as it is found that bilingual programs are successful at increasing years of schooling, academic achievement and reducing outcome gaps (Lesiewski, 2015). However, these studies have typically been confined to small geographical areas and specific schools. For instance, Hynsjo & Damon (2016) find positive results for education outcomes for indigenous children in a single school in Peru. However, no large scale or nation-wide bilingual programs have been evaluated. As a result, questions may be raised as to whether ASEP may in fact be as successful at improving education outcomes, particularly given Bolivia's precarious situation and lack of resources. Is there a disadvantage to the implementation of bilingual programs across the nation? Can ASEP serve as a model for other multilingual developing countries? Ultimately, this begs the question: to what

extent has bilingual education increased educational outcomes in Bolivia?

This investigation answers the question by empirically analyzing how education outcomes have fared in Bolivia. In particular, it uses a difference-in-difference methodology, which exploits the fact that some areas used indigenous languages for instruction while others stuck to Spanish, to look at the change in outcomes following the reform. The results reveal that bilingual education was indeed successful at increasing educational outcomes: the control group saw an increment in years of schooling, literacy, numeracy and primary schooling by 0.514 years, 8.6, 8.7 and 10.9 percent,

(6)

5

respectively. In addition, females, rural and indigenous individuals, who have traditionally been marginalized groups, also saw big improvements in their educational outcomes. These results would corroborate the claims of the pedagogical literature and indicate that the expansion of bilingual education into large scale programs remains successful at improving education outcomes. It is also plausible that the efficiency gains from bilingual education are vast. In fact, Patrinos & Velez (2009), estimate that a bilingual program in Guatemala can result in cost savings arising from reduced repetition alone equal to the cost of primary education for 100,000 children. Given that Bolivia’s program is larger in size and scope, it is very likely that through economies of scale the cost savings and benefits accrued outweigh those found previously. This is explored in greater detail later in Section VIII.

The remainder of this paper is organized as follows. Section II presents a brief overview of Bolivia’s education situation and describes the reform. Section III outlines a literature review on education interventions and, specifically, bilingual education. Section IV enumerates the investigation’s hypotheses. Section V depicts the data used. Section VI details the methodology employed. Section VII presents the investigation’s results before concluding in Section VIII.

II. Context

Education in Bolivia is characterized by poor access and low quality, creating a bleak situation for the Bolivian people who are often some of the most poorly educated in Latin America. Despite some progress in recent years, which has seen primary education rates increase over time, dropout rates are rampant, over-age students are becoming a growing trend and illiteracy rates continue to be high. Much of this is the direct consequence of the poor educational quality provided in Bolivian schools. In addition, the poor infrastructure and lack of resources devoted to education have exacerbated the problem (World Bank, 2017).

These shortcomings in education affect indigenous people in rural areas the most, who also tend to be the poorest and most vulnerable. Official statistics reveal that rural areas still see illiteracy rates at 26 percent, while 43 percent of indigenous women are unable to read and write (INE, 2012). Separately, some scholars have estimated that functional illiteracy amounts to 70 percent in rural areas. School dropouts and grade repetition are also big issues. Scholars claim the repetition rate is at 38 percent, with children only attaining 2.2 grades for 4.8 years enrolled in primary school (Patrinos & Velez, 2009).

Much of this may be explained by the impoverishment affecting indigenous people and rural areas, where the lack of provision of social services, the isolated geographical location and discrimination have placed indigenous children at an extreme disadvantage. In fact, poverty in

(7)

6

indigenous areas is the highest in the nation, peaking at 80 percent for children under five. In addition, the lack of access to quality health and nutrition services have increased vulnerabilities among poor and rural populations. In 2003, for instance, chronic malnutrition affected one third of Bolivian children under five, 42 percent of which lived in rural areas (World Bank, 2017). The shortages hitting indigenous areas is compounded by low quality, severely underfunded and insufficient education institutions. Some areas, for instance, lack a university, children are unable to continue their education through to high school, and some schools even lack running water (Storm, 2011). This has created a lack of trust in education institutions both by parents and students. In fact, Appendix 1 presents evidence revealing that many children are often pulled out of school by their parents to attend work and help in house chores. It is thus no surprise that indigenous children are more likely to work, be illiterate, repeat grades and quit school early. Similarly, women are largely marginalized when it comes to education. As shown in Table 1, women are comparatively worse off than men, while indigenous rural women register extremely low education outcomes before the reform.

Table 1: Comparison of Men’s and Females’ Education Levels Pre-Reform

_SchoolingYears Literacy Numeracy School Completion

Primary Secondary

Urban Non-Indigenous Men 7.26 84% 82% 81% 56% Urban Non-Indigenous

Women 6.22 81% 80% 72% 48%

Urban Indigenous Women 6.09 74% 72% 51% 29% Rural Indigenous Women _5.73 _68% _66% _26% _10%

Notes: Urban non-indigenous men register the highest education attainment levels in Bolivia, hence their education levels are included to serve as a benchmark for comparison; percentages are rounded to nearest whole number; years of schooling

are rounded to the nearest 2 decimal places. Source: INE (2012)

In addition, Bolivia’s historical context also is largely responsible for this bleak situation. Bolivia’s racist and colonial past institutionalized segregation that extended to education. Before the 1952 Revolution, indigenous people were excluded from civil society, including education. Following the Revolution, the 1955 Education Act guaranteed every citizen access to education. However, this implied access to a school based on the language and ruling culture with the explicit aim of “assimilating” and “civilizing” indigenous people. As a result, in 1900, 85 percent of the entire population was illiterate; in 1960, 70 percent was still unable to read or write. These effects have endured over time so that today Bolivia continues to lag behind in international comparisons of academic achievement and has one of the highest worldwide internal inequalities in performance (Garcia de Rosa et al., 2015). Thus, the poor efforts combined with the disregard for indigenous people resulted in incomplete schooling and poor distribution and content that have been prevalent ever since.

(8)

7

To overturn this, recent Bolivian governments have implemented a series of reforms over the years. Some have involved cash transfer programs at the micro level, while others have been large scale education reforms such as the 1994 Education Reform that improved curricula, teacher training and decentralized funding. Yet, the most ambitious, innovative and large-scale effort came with the change in government that witnessed Morales’s rise to power in 2006. In 2006, the Morales government implemented the Avelino Siñani Elizardo Pérez (ASEP) law that was instituted to bring about “decolonization, intra and interculturalism and plurilingualism” (Lopes, 2011). An official government description of “decolonization” of education may be found in Appendix 2. However, ASEP’s main cornerstone is the implementation of “intercultural bilingual education” (IBE). This established that each department should use two official languages in its schools for instruction. One is Spanish, and the other is decided considering the use, convenience, and circumstances of the population as a whole or the territory in question (Jimenez Quispe, 2011). For example, Potosí, an area that includes a large Quechua indigenous population, utilizes Spanish and Quechua in schools. Beni, on the other hand, uses primarily Spanish as its indigenous population is rather small. These are depicted below in Table 2 and illustrated in Appendix 3. Nevertheless, the IBE program emphasizes that while indigenous languages are to be used heavily, particularly in the lower grades of primary school, eventually all children are expected to be proficient in Spanish. ASEP covers both primary education, which lasts 6 years and is compulsory, and was implemented with immediate effect (Talavera-Simoni, 2013). Although ASEP affects all education in Bolivia, private schools are exempt as they do not follow the public education program. Although this would make private school students a potential alternative control group, the lack of sufficient observations in the data make this task impractible.

Bolivia thus follows in the footsteps of many developing countries that implement bilingual education programs. Particularly in Latin America, the popularity of bilingual education has increased in recent years, especially in countries with large indigenous populations. In Guatemala, for instance, efforts to develop programs aimed at the education of indigenous children began as early as 1930. Today, Guatemala implements one of the most rigorous and complete bilingual education programs -PRONEIB - that even includes scholarships for overseas training of technical staff (Enge et al., 1996). Similarly, Peru instituted bilingual education during the 1970s, expanding to over 1,200 schools today (Hynsjo et al., 2016). However, Bolivia’s experience represents the largest effort to implement bilingual education at the nation-wide level.

(9)

8

Table 2: Official Languages Used for Instruction per

Department in Bolivia Post-Reform Region Name Languages

Chuquisaca _{Spanish, Quechua & Aymara} La Paz _{Spanish, Quechua & Aymara}

Cochabamba _{Spanish, Quechua, Yuracare, Yuqui &} Aymara

Oruro _{Spanish, Quechua & Aymara} Potosi _{Spanish & Quechua}

Tarija _{Spanish, Quechua, Aymara & Mataco} Santa Cruz _{Spanish, Quechua & Guarani} Beni _Spanish

Pando _{Spanish, Quechua & Aymara}

Notes: Reform was passed in 2006. Source: Crevels & Muysken (2009)

This has also come at a cost, often manifested in challenges to implementation due to the complex processes involved in such large-scale reforms. Most notably, a lack of bilingual education resources has been documented (Talavera-Simoni, 2013). In addition, teacher and popular resistance may have delayed and threatened implementation (Cardozo, 2012). Moreover, the big change in teaching methods have posed difficulties for teachers when having to adapt to the new guidelines (Talavera-Simoni, 2013). Other challenges to implementation, their implications for the investigation and how they are accounted for are dealt with more extensively in Section IV.

III. Literature Review

Over the years, much attention has been devoted towards studying education in developing countries, resulting in the establishment of a large literature that investigates the impact of several interventions on schooling and education achievement. Traditionally, researchers have looked at increasing the quality and availability of schooling to improve the education outcomes of children across the developing world. This has led to a range of price-reducing interventions such as abolishing school fees, building more schools, and providing families with targeted cash or in-kind transfers, some of which have been quite successful at increasing schooling and education outcomes (Marshall, 2011). However, the effectiveness of these efforts to really achieve increased education and learning has been put under question. In fact, it has been widely documented that while more children in developing countries are attending school for more years, they are not learning. The failure to achieve significant improvements here may be explained in large part due to the widespread inefficiencies present across

(10)

9

schooling systems in the developing world and because these interventions are not tailored to children’s needs, which have further exacerbated the already disadvantaged position of the poor and marginalized.

The inability to find a solution to this dismal situation has prompted scholars to look for alternatives by examining how learning is conducted. Here, language of instruction has gained considerable momentum as most formal education in developing countries occurs in the language of their ex-colonizers, which often remains foreign to a significant part of the population. Therefore, education in the language of ex-colonizers often results in inefficient learning, wasteful and costly educational systems, particularly in multilingual, plurinational contexts. These facts have thus inspired a new strand of research that looks at language of instruction and led to the implementation of bilingual education programs to better cater to children’s learning needs.

Examining the impact of these bilingual programs is an increasing body of evidence, much of which is of a pedagogical and qualitative nature. These studies often emphasize the real and potential benefits of bilingual programs from a didactive perspective. For instance, Benson (2000) reveals how bilingual programs provide content area instruction (e.g. mathematics) in a language that children understand. This prevents instruction from being postponed until mastery of the foreign language, accelerating learning. Similarly, bilingual education programs use mother tongue to teach beginning reading and writing, facilitating the understanding of both sound-symbol correspondence, particularly for phonetic languages, and the connection between written and spoken communication. In addition, instruction in the mother tongue allows for greater interaction between students and teachers. Hornberger & Chick (2001) show this by illustrating how in cases when schooling is in the foreign language teachers are often only able to elicit rote responses. Instead, when education is in the mother tongue, teachers can interact more naturally with students, improving learning and affective considerations (Baker, 2011). Furthermore, this may allow teachers to get a better understanding of what their students are learning and thus adjust their teaching, which has been corroborated in d’Emilio (2001). Moreover, the pedagogical perspective stresses that the benefits of bilingual education can occur outside the realm of schooling. For example, Dalby (1985) documents how bilingual education programs have positive impacts for affective factors such as motivation and self-esteem. Dutcher & Trucker (1996) also show how bilingual schooling can result in increased pride for home languages and culture and has the potential to bolster political participation and social relations. Thus, the widely documented and potential benefits have led to the wide acceptance and seldom questioning of bilingual education on pedagogical grounds other than to refine basic premises and point the challenges each program faces given the country context.

These claims have been examined in the empirical literature, with some studies questioning the pedagogical perspective. For example, studies looking at bilingual programs in developed countries, particularly those of the US targeting native Spanish speakers, occasionally reject the pedagogical perspective. Willig (1985) finds that participation in bilingual programs in the US produced only small

(11)

10

to moderate positive results for tests of reading, language and mathematics. Similarly, Curiel et al. (1986) show that differences in GPA between bilinguals and non-bilinguals in Houston were marginal, albeit bilinguals being less likely to dropout and experience retentions. However, Greene’s (1998) meta-analysis of scholarly research contradicts these studies. First, he highlights their methodological shortcomings by claiming that many studies lack a control and treatment group that have been randomized. Second, his review reveals that children in the US with limited English proficiency who are taught using at least some of their native language perform significantly better on standardized tests than similar children who are taught only in English. These contradictions have sparked lively debates on bilingual education in the US.

Nevertheless, the evidence emanating from developing countries does support the notion that bilingual programs may have positive results for education outcomes. For example, Walter & Dekker’s (2011) Lubuagan study tests the impact of a bilingual program in a rural area in the Philippines that taught reading and writing and other key subjects in the local language while Tagalog and English were taught as separate subjects. The results found that end of year standardized testing resulted in bilinguals scoring 20 to 25 percentage points higher than non-bilinguals. Similarly, Hovens (2002) looks at bilingual programs in Niger and Gambia and demonstrates that pupils who started education in their mother tongue could read and write better even in their second language. In addition, bilingual classrooms were also more stimulating, interacting and relaxed, and those who gained most from participation were rural children and girls. Benson (2000) instead examines the impact of PEBIMO, a bilingual program in Mozambique that utilized two Bantu languages for instruction in lower primary schooling. Evaluations done in the final two years of the experiment reveal that students benefited greatly in terms of biliteracy and bilingualism. Specifically, PEBIMO achieved 3 percent lower dropout rates than the national average and significantly higher passing rates by 10 to 30 percent. Qualitative evidence further reveals that students saw improvements in self-confidence, self-esteem and classroom participation.

In addition, the literature emphasizes how bilingual programs have been successful at decreasing learning gaps and reinforcing the position of the marginalized, which has often been tested in the context of Latin America and their indigenous populations. This is revealed in Hynsjo & Damon (2016) who find that indigenous children who attend Quechua-medium schools in Peru achieve 0.54 SD higher mathematics scores than indigenous children who attend Spanish-medium schools. Their results lead them to conclude that indigenous-language medium instruction can play a significant role in ameliorating the indigenous test score gap. Similarly, McEwan (2008) finds bilingual education programs in Chile result in increased test scores for indigenous students. Pascharopolous (1993) adds to this by showing how bilingual education is a good investment for indigenous people to address their lower levels of education achievement and labor market rewards. Likewise, there is ample evidence showing how the implementation of bilingual education greatly improves opportunities for educational

(12)

11

access and attainment for female students. Sichra (1992), for instance, finds that girls who learn in their mother tongue stay in school longer, are more likely to be identified as good students, do better in achievement tests and repeat grads less often than their peers who do not get home language instruction. Benson (2005) seconds these claims by noting that it is the learner’s mother tongue that holds the key to making schooling more inclusive for all disadvantaged groups.

However, one may come across important limitations in the literature. Firstly, very scant quantitative evidence exists on the impact of bilingual education, particularly for developing countries. This, combined with the conflicting evidence often found in developed country studies, reveals that there is insufficient evidence to conclude in what direction bilingualism points to (Baker, 2011). Second, most studies analyze small-scale programs that are typically confined to small geographical areas or small subpopulations, as in Hynsjo & Damon (2016). As a result, there is a gap in the literature that fails to address the impact of large-scale, nation-wide bilingual programs in developing countries. Although this is likely the direct result of programs being designed to be in such a way, evaluating a large-scale program is essential to address the benefits and challenges it may encounter, particularly in developing countries where scarcity of resources is most acute (Benson, 2005).

Thus, the examination of the literature reveals that this investigation is valuable for several reasons. First, it contributes to the literature on bilingual programs in developing countries for which there is scant and inconclusive quantitative evidence. Second, it looks at a nation-wide, large-scale policy, which has been missed in previous evaluations. Third, analyzing Bolivia, which has a dismal education situation, a large indigenous population and acute scarcity of resources, provides an excellent context to examine the claims previously outlined.

IV. Hypotheses

The claims and evidence put forward in the literature facilitate inferring several hypotheses predicting how bilingual education will impact children’s education outcomes. In the case of Bolivia, it seems plausible that this impact will occur through two main channels. First, bilingual education may result in increased demand for education, both from parents and children. Parents may see the new education methods as being more practical, more relatable and hence more useful for later in life. This would lead them to value school quality more, adjusting their preferences so that they allocate more resources to their child’s schooling (Glewwe, 2002). In addition, the implementation of bilingual programs may bolster participation in society, which can translate into increased willingness to send children to school and have them attend for longer. This is especially applicable to Bolivia’s context where the indigenous and rural population has felt alienated, particularly when it comes to education (Howard, 2009). Thus, parents may largely favor bilingual programs, which can result in a greater willingness to send their child to school for longer. Similarly, children may also experience a greater

(13)

12

willingness to attend school. This could be the result of the role-model effect, which engages students’ effort, confidence and enthusiasm(Dalby, 1985). Moreover, children may become more motivated as they are now better able to understand the material, become more conscientious and want to attend and stay in school for longer. Furthermore, the expected increase in demand for schooling from both parents and children may impact dropout rates through several channels. Since both are now more willing to send and attend school, combined with a higher degree of conscientiousness, dropout rates may be expected to decrease. The expected increase in schooling, better attitudes and self-esteem may also lead to more success in school, making exiting from school less likely (Curiel et al., 1986). Additionally, the greater perceived value placed on school by both parents and children may deter pulling children out of school in order to attend household chores and work. As a result, the below hypothesis may be expected.

Hypothesis 1: Bilingual education in Bolivia will have a positive impact on schooling, resulting in an

increase in children’s years of schooling. Similarly, bilingual education will result in a negative impact on dropout rates that will be evidenced by more school completion.

Second, it is plausible that bilingual education in Bolivia constitutes an improvement in teaching methods, which in turn may produce better learning outcomes directly and through its interaction with greater schooling (Fuller & Clarke, 1994). Should higher demand for schooling be attained, manifested in more years of schooling and less dropouts, children will be able to learn for longer. Longer time learning combined with better teaching methods and the expected psychosocial benefits outlined previously can result in enhanced learning outcomes (Patrinos & Velez, 2009). In addition, bilingual education can improve teaching efficiency and enhance learning as it facilitates content area instruction and improves the understanding of written and spoken communication (Benson, 2000). Furthermore, it may result in greater interaction between teachers and students, allowing teachers to assess students’ progress and needs. Greater interaction can also foster cognitive stimulation, further bettering learning outcomes (World Bank, 2017). Because of these interactions, it is plausible that improved teaching methods will culminate in following hypothesis.

Hypothesis 2: Bilingual education will have a positive impact on learning outcomes that will be

reflected in greater numeracy and literacy skills.

Finally, it is likely that ASEP will impact traditionally marginalized groups in a positive way. The literature emphasizes how bilingual programs have been successful at decreasing learning gaps, particularly in the context of Latin American countries and their indigenous and rural populations. This is revealed in Hynsjo & Damon (2016) who find that indigenous children who attend Quechua-medium schools in Peru achieve 0.54 SD higher mathematics scores than indigenous children who attend

(14)

13

Spanish-medium schools. Similarly, Awopetu (2016) finds that girls who attend bilingual schools attain more years of schooling, improved learning outcomes, more school completion and less dropouts. Therefore, the use of bilingual instruction can play a significant role in ameliorating the gaps between advantaged and marginalized groups, hence making Hypothesis 3 plausible.

Hypothesis 3: Bilingual education in Bolivia will result in positive outcomes for traditionally

marginalized communities: females and rural individuals.

V. Data

The data utilized for the investigation was retrieved from Bolivia’s National Institute for Statistics (INE, 2017). The survey being looked at was carried out in 2017 and provides information on living conditions through data collection on subjects’ health and health care access, fertility, employment, income, expenditures, basic services and housing. Most importantly for the investigation, the data reveals subjects’ year of birth, schooling, education attainment and life outcomes. As such, the data reveals information on the dependent and independent variables and allows controlling for various possible confounding factors. The sample size is 38, 201 individuals.

Table 3 presents descriptive statistics of key variables being examined for the full sample and the control and treatment groups of Beni and Potosi, respectively. This includes the main dependent variables of literacy, which takes the value of 1 if the individual can read and write and 0 otherwise; numeracy, taking the value of 1 if the individual can perform addition and subtraction and 0 otherwise; and school completion, taking the value of 1 if the individual (previously enrolled) has completed primary or secondary schooling and 0 otherwise. Average years of schooling for the full sample and control and treatment groups also appear. Overall these are relatively low, albeit higher than in the control and treatment areas, with Beni enjoying 0.309 more years of schooling than Potosi. This pattern repeats itself for literacy and numeracy, which are lower in the control and treatment areas than the full sample. Comparing Beni versus Potosi reveals the former reports 0.064 and 0.062 more literacy and numeracy, respectively. It is therefore no surprise that school completion is poor in both areas. While primary school completion is 0.009 points greater in Beni than Potosi, these differences increase to 0.017 points when looking at secondary school completion. The p-Values reported for these differences reveal they are not significant. Additional summary statistics for the outcome variables may be found in Appendix 4.

(15)

14

Table 3: Descriptive Statistics

Full Sample Beni Potosi

Variable Obs. Mean Std. Dev. Obs. Mean Std. Dev. Obs. Mean Std. Dev. Difference p-Value Age 38,201 29.046 20.830 2,008 30.419 22.457 1,824 30.438 23.413 -0.019 0.421 Female 38,201 0.507 0.500 2,008 0.507 0.500 1,824 0.536 0.499 -0.029 0.861 Indigenous 38,201 0.287 0.452 2,008 0.087 0.281 1,824 0.569 0.495 -0.482 0.000 Quechua 38,201 0.130 0.337 2,008 0.020 0.140 1,824 0.542 0.498 -0.522 0.000 Aymara 38,201 0.134 0.341 2,008 0.014 0.117 1,824 0.008 0.087 0.006 0.000 Guarayo 38,201 0.000 0.018 2,008 0.001 0.032 1,824 0.000 0.000 0.001 0.348 Guarani 38,201 0.006 0.074 2,008 0.000 0.022 1,824 0.001 0.023 -0.00005 0.068 Rural 38,201 0.212 0.409 2,008 0.460 0.424 1,824 0.480 0.500 -0.020 0.199 Years Schooling 35,104 8.161 5.473 1,790 7.659 5.039 1,683 7.351 5.326 0.309 0.185 Incomplete Primary Schooling 35,104 0.333 0.471 1,790 0.355 0.479 1,683 0.364 0.481 -0.009 0.582 Completed Primary Schooling 35,104 0.667 0.471 1,790 0.645 0.479 1,683 0.636 0.481 0.009 0.582 Completed Secondary Schooling 35,104 0.364 0.481 1,790 0.309 0.462 1,683 0.292 0.455 0.017 0.164 Numeracy 38,201 0.803 0.397 2,008 0.736 0.420 1,824 0.707 0.455 0.028 0.596 Literacy 38,201 0.817 0.386 2,008 0.751 0.414 1,824 0.719 0.450 0.033 0.267 Public School 13,139 0.869 0.337 681 0.9325 0.010 635 0.9317 0.009 0.0008 0.568 Breakfast 10,341 0.791 0.406 579 0.927 0.261 558 0.959 0.199 -0.032 0.381 Bono Jacinto Pinto 10,341 0.733 0.443 579 0.753 0.434 558 0.769 0.422 -0.016 0.533 Household Income 38,201 5157.161 4522.516 2,008 4202.600 3443.770 1,824 4077.262 4680.297 125.338 0.991 Below Poverty Line 38,201 0.350 0.477 2,008 0.409 0.492 1,824 0.439 0.496 -0.030 0.186 Below Extreme Poverty Line 38,201 0.150 0.357 2,008 0.195 0.396 1,824 0.274 0.446 -0.079 0.584 Owns Phone 38,201 0.624 0.484 2,008 0.541 0.498 1,824 0.581 0.493 -0.040 0.235 Internet Access 38,201 0.349 0.477 2,008 0.278 0.448 1,824 0.275 0.446 0.003 0.205 Owns Computer 38,201 0.257 0.437 2,008 0.108 0.310 1,824 0.186 0.389 -0.078 0.645

Note: p-Values are based on t-tests

These values are comparatively lower than those of the rest of Latin America. Mean values for years of schooling, literacy and numeracy are some of the lowest in the region, while primary school completion is only lower in El Salvador, Honduras and the Dominican Republic (da Rosa et al., 2015). Similarly, education achievement follows much of the same pattern. Furthermore, although data on

(16)

15

school performance levels is not available for Bolivia, it is plausible that learning levels are far from satisfactory and trail behind region averages (Infante et al., 2013).

Descriptive statistics for the control variables and their differences are also reported in Table 3. This reveals key information on indigenous status, which takes on the value of 1 if the individual self-identifies as indigenous and 0 otherwise. Similarly, the variables Quechua, Aymara, Guarani and Guarayo reveal information on the specific indigenous group the individual belongs to. Information on the individual’s sex, age, area and socioeconomic status is also specified. No significant differences are found between the groups for most variables. However, the treatment group has a much greater indigenous, predominantly Quechua, population than the control group. This is largely expected given the concentration of indigenous people in Potosi and is therefore controlled for in the regressions. Breakfast reveals whether a child receives breakfast at school and Jacinto Pinto whether a child receives a government cash transfer providing a stipend for school attendance. These are included as the policies were passed roughly at the same time as ASEP and therefore could influence the results (Bebbington et al., 2018). However, these are later dropped from the regressions due to a lack of enough observations. Other interesting points worth noting are the differences for the socioeconomic control variables. Household income and internet access are greater in Beni than Potosi, while the opposite is true for phone use and computer use. These differences, which are not significant, are controlled for in the regressions.

VI. Methodology

i. Empirical Strategy

To estimate the effects of bilingual education on educational outcomes, the investigation follows much of the literature and uses a difference-in-difference framework. In particular, the investigation exploits the fact that some areas are home to a larger indigenous population than others and thus utilized indigenous languages as medium of instruction post reform while others stuck to Spanish (Jimenez Quispe, 2011). In addition, it exploits that individuals who entered primary school after 2006 (after the reform was passed) were exposed to the reform, while those who completed primary schooling before 2006 were not. Therefore, in Beni, which has a small indigenous and predominantly Spanish speaking population, Spanish continued to be utilized as language of instruction after 2006. Conversely, Potosi’s large Quechua population meant schools took up Quechua for

(17)

16

instruction. These differences in language of instruction according exposure allow for Beni to serve as the control group and for Potosi to serve as the treatment group. This is illustrated in Table 4.

Table 4: Languages Used in Control and Treatment Groups Pre

and Post Reform

Pre-2006 Post-2006 Beni Spanish Spanish Potosi Spanish Spanish & Quechua

Notes: Reform was passed in 2006; Pre-2006 refers to individuals who were not exposed to the reform; Post-2006 refers to individuals who were exposed

to the reform. Source: Jimenez Quispe (2011)

Furthermore, Beni and Potosi were selected as the control and treatment areas as the common trend assumption between both groups is likely to hold. This is illustrated in Figure 1, which depicts how the evolution of years of schooling for both areas was roughly equal before the reform. Therefore, the DID requirement of constant differences between treatment and control groups over time is satisfied. Common trend analysis for the rest of outcome variables, depicted in Appendix 5, further confirms the existence of a parallel trend. Other areas are excluded from the analysis primarily because they often include several indigenous groups, which may difficult controlling for changes in the language of instruction utilized.

One possible complication that arises when assigning the control and treatment groups is the availability of data only on where the individual lives at the time of the survey, not where they were born or attended primary school. This may raise estimation issues if, for instance, individuals were born in Beni and then later migrated to Potosi. This could cause biased results as individuals would be wrongfully assigned to treatment or control groups. However, Andersen (2002) claims interstate migration in Bolivia is rather low, particularly between Highlands (Potosi) and Lowlands (Beni). Therefore, it is assumed that individuals attend school in the department they reside and are thus correctly assigned to the control or treatment group.

(18)

17

The age of the subjects at the time of the census is used to define whether an individual was exposed to the education reform or not. Those individuals who completed primary school before 2006 were not exposed to the reform and were thus instructed in Spanish irrespective of their location. Given that primary school in Bolivia lasts 6 years and starts roughly when a child is 6 years old, the Not Exposed group would be older than 25 at the time of the census in 2017. The group of individuals who were fully exposed to the reform were those who attended primary school and instructed in bilingual education. Therefore, they started primary school after once ASEP was passed in 2006, hence being below the age of 18 in 2017.

A third group of individuals, those that were partially exposed to the reform, may be further defined. These are students who were still attending primary school at the time of the reform but had not finished. There are several possibilities on how to deal and define this group. On the one hand, should the implementation of ASEP be gradual, the Partly Exposed group would not be affected by the reform at all, as it would only concern those individuals that started primary school after 2006. Therefore, the Partly Exposed would continue to be instructed in Spanish, hence actually belonging to the Not Exposed group. On the other hand, if the take up of bilingual education would have been immediate, the Partly Exposed group would have had a share of their primary school experience instructed in Spanish and the remaining share being bilingual. To deal with this, a value relating the share of primary school attended after the reform is attributed to each Partially Exposed observation (e.g. should a child attend 4 out of 6 years of primary school after ASEP is passed, his exposure share is 0.667). Although not much evidence exists regarding the process of implementation, it is plausible that, given the radical Morales agenda, reform take up was rather immediate (Cardozo et al., 2013) and

Figure 1: Mean Years of Schooling per Age Group Pre-Reform

Notes: Age Group refers to the age bracket individuals conform to. Age brackets are the following: Group 8: 30-34, Group 7: 35-39, Group 6: 40-44, Group 5: 45-49, Group 4: 50-54,

(19)

18

therefore the latter option is preferred. The former option as well as dropping the Partly Exposed group is included in the regressions as robustness checks (Seid, 2016). A further assumption made is that all individuals attend the correct grade for their age, implying late enrollments and grade repetition is discounted. Although this may seem as implausible given Bolivia’s context, it must be accepted as it cannot be refuted.

ii. Regression Specification

To estimate the impact of bilingual education on education outcomes following DID approach, the outcome variable of interest is regressed on the dummy variables for Treatment and Exposure, as well as their Interaction Term, which is the coefficient of interest. Therefore, the following specification is estimated

(1)

Where Yits is the outcome variable of interest. To test the hypotheses formulated in Section IV,

the outcome variables estimated are Years of Schooling, Literacy, Numeracy and Primary School Completion. Xits refers to a vector of control variables added to increase the precision of the estimates.

These variables are specified in Table 2. However, the Breakfast dummy and the Jacinto Pinto cash grant dummy are dropped from the specification due to a lack of sufficient observations. Finally, eits refers to the random error term. The regression is then performed separately to split the results by female, rural and indigenous individuals. This is to reveal whether bilingual education has been successful in empowering and benefitting those who have traditionally been marginalized, thus testing Hypothesis 3.

iii. Estimation Issues

Apart from the potential sources of bias discussed previously, the results may be influenced through several other sources that produce biased estimates. One concern arises from the fact that attitudes towards President Morales and his agenda may influence the way the ASEP reform is perceived. If individuals base their judgement of ASEP on their positive (negative) attitudes towards the Morales government, this may impact education outcomes in a way that is not strictly based on the way the new education landscape is perceived. For instance, if individuals feel empowered or identified with the new government they may become more involved in society as they feel more empowered or

(20)

19

identify with the new regime as opposed to the way they value changes in institutions. This is an important concern given the popularity of the Morales government, particularly amongst the indigenous population (Postero, 2010). However, the lack of qualitative evidence found here may evince that this issue is not too concerning. In fact, Appendix 6reveals that the reception of ASEP has been mixed, even amongst indigenous groups. Thus, it seems that identity politics do not seem to play a big role, hence this source of bias is not concerning.

A second concern, which is raised in Hynsjo & Damon (2016), deals with parents’ attitudes towards bilingual education. According to Hynsjo & Damon, should parents express disapproval of bilingual programs, possibly because they fear it may hamper their child’s acquisition of Spanish, they may migrate to other areas where instruction is in Spanish. Yet, Lopez (2002) shows how this is largely not the case in developing countries implementing Bilingual education. Furthermore, the lack of evidence in Bolivia’s context seems to indicate it is not a pressing issue.

Finally, there is the possibility that bilingual education is not fully implemented in practice or encounters serious limitations to implementation. Should this be the case, the coefficients for the estimated variables would arise due to other effects and not the impact of bilingual education, thus severely hampering the validity of the results. This could be plausible given Bolivia’s poverty-stricken state, which translates in a lack of resources, including those devoted towards education (World Bank, 2017). As a result, implementation of the reform may be slowed or even halted in some areas. In addition, the quality of bilingual education may slacken if efforts are not channeled correctly. In Bolivia’s case, Talavera-Simoni (2013) has documented a lack of bilingual education resources, while Cardozo (2012) records teacher and popular resistance. However, since ASEP’s approval, government efforts have been devoted towards teacher training and quality control, thus allowing for the assumption that school quality, particularly that of bilingual schools, will be significantly hampered (Talavera-Simoni, 2013).

VII. Results

i. Baseline Results

Table 5 presents the baseline results after performing the regressions specified in Equation (1) when partly exposed individuals are included in the analysis. Results for each specification, with and without the vector of control variables Xits, is also included. As revealed from the coefficient of interest,

(21)

20

increase of 0.514 years of schooling. This result is robust to the inclusion of control variables. The increase in years of schooling reported is largely corroborated throughout the empirical evidence and pedagogical literature and explained by an increase in the demand for education (Patrinos & Velez, 2009).

Table 5: Baseline Regression Results

Variable Years Schooling Literacy Numeracy Primary School Completion Interaction 0.514*** 0.437*** 0.086*** 0.047*** 0.087*** 0.066*** 0.109*** 0.073*** (0.018) (0.015) (0.017) (0.014) (0.017) (0.013) (0.016) (0.010) Treatment 1.022*** 1.866*** 0.182*** -0.063*** -0.180*** -0.068*** -0.237*** -0.143*** (0.259) (0.214) (0.012) (0.010) (0.012) (0.010) (0.020) (0.009) Exposed 4.793*** 0.322 -0.259*** 0.369*** -0.277*** 0.440*** -0.333*** 0.525*** (0.197) (0.321) (0.018) (0.037) (0.017) (0.038) (0.016) (0.040)

Xits Included No Yes No Yes No Yes No Yes

Control Variables included: age, sex, rural, public school, household income, owns phone, internet access, owns computer; Standard Errors appear in parentheses;

p<0.1*, p<0.05**, p<0.01*** Numbers are rounded to 3 significant figures

The results also show a positive impact on literacy and numeracy. The Interaction coefficient reveals that ASEP has caused an increase in literacy of about 8.6 percent, while numeracy increased by 8.7 percent. These results are also robust to the inclusion of controls Xits. Improvements in literacy and

numeracy could be explained by the increases in teaching and learning efficiency arising from improved sound-symbol correspondence, better student-teacher interactions and accelerated content area instruction. Benson (2000) and Hornberger & Chick (2001) further corroborate these results.

Finally, Table 5 reveals the results for primary school completion. The coefficient of interest shows that primary school completion has risen by 10 points since the inception of ASEP. This would indicate that school dropouts have decreased as more individuals who are enrolled complete primary schooling. Explanations for these results may be found in enhanced academic achievement and psychosocial skill development that improve student well-being, motivation and self-esteem, all of which improve school dropouts and school completion (Prevatt & Kelly, 2003). Therefore, the evidence gathered supports Hypotheses 1 and 2.

As explained previously, the results are then verified by redefining Partially Exposed individuals (Seid, 2016). First, the observations for Partially Exposed individuals are dropped and excluded from the analysis. The regression results for this test are presented in Table 6. As shown, the size and direction of the coefficients remain relatively similar to those presented in Table 4. In addition, except for a few coefficients, they remain largely significant. Therefore, the results conferred in Table 6 ratify those of Table 5. Second, Partially Exposed individuals are defined as Not Exposed. The regression results for this method are shown in Table 7, which indicate contrasting coefficient values. Furthermore, the significance of the coefficients has largely decreased, while the Interaction Term

(22)

21

coefficients are now insignificant. Therefore, the available evidence confirming the reform was implemented immediately (Talavera-Simoni, 2013) together with the insignificant results lead to the preference of the inclusion of Partially Exposed individuals as in the regressions previously conducted.

Table 6: Regression Results – Partially Exposed Dropped

Variable Years Schooling Literacy Numeracy Primary School Completion Interaction 0.474*** 0.398*** 0.072*** 0.016 0.099*** 0.015 0.153*** 0.083*** (0.054) (0.046) (0.019) (0.012) (0.027) (0.0127) (0.042) (0.014) Treatment -2.498*** -2.040*** -0.180*** -0.018* -0.177 *** -0.168*** -0.248*** -0.097* (0.435) (0.381) (0.025) (0.002) (0.014) (0.016) (0.062) (0.018) Exposed .574*** -3.057*** 0.191*** 0.004 0.199*** 0.317*** -0.276*** 0.004 (0.129) (0.116) (0.058) (0.003) (0.029) (0.012) (0.026) (0.011)

Table 7: Regression Results – Partially Exposed as Not Exposed

Variable Years Schooling Literacy Numeracy Primary School Completion Interaction 0.933 0.927 0.016 0.007 0.032 0.018 0.050 0.071 (0.731) (0.920) (0.082) (0.015) (0.036) (0.714) (0.181) (0.239) Treatment 0.247 -1.129* -0.053*** -0.065*** -0.106*** -0.066* 0.103 -0.052 (0.358) (0.147) (0.003) (0.010) (0.038) (0.033) (0.251) (0.026) Exposed -0.922 0.569 -0.462*** 0.278*** -0. 925*** 0.474*** 0.503*** -0.068*** (0.703) (0.869) (0.041) (0.020) (0.190) (0.032) (0.062) (0.005)

ii. Results for Females, Rural and Indigenous Individuals

Table 8 presents the results for the regression specified in Equation (1) after splitting the observations by sex. As shown, bilingual education has resulted in increased years of schooling, literacy, numeracy and primary school completion for females. The coefficients for the interaction term, all of which are significant, would thus confirm that females have benefited greatly from the reform and have improved their disadvantaged position. This is largely corroborated across the literature as in

(23)

22

Sichra (1992) who finds that girls who learn in familiar languages stay in school longer, are more likely to be identified as good students, do better in achievement tests and repeat grades less often than their peers who do not get home language instruction.

Table 8: Regression Results - Females

Variable Years Schooling Literacy Numeracy Primary School Completion Interaction 0.722*** 0.417*** 0.091*** 0.062*** 0.095*** 0.065*** 0.117*** 0.059*** (0.111) (0.026) (0.019) (0.015) (0.018) (0.014) (0.020) (0.011) Treatment -2.744*** -0.617*** -0.204*** -0.065*** -0.205*** -0.066*** -0.235*** -0.112*** (0.236) (0.134) (0.015) (0.010) (0.015) (0.010) (0.022) (0.009) Exposed -3.917*** -0.599*** -0.208*** 0.278*** -0.224*** 0.317*** -0.264*** 0.351*** (0.218) (0.126) (0.028) (0.020) (0.029) (0.012) (0.024) (0.021)

Rural individuals have also reaped the benefits of the reform as evidenced in Table 9. The Interaction coefficients for years of schooling, literacy, numeracy and primary schooling are largely significant and higher than for all the other groups. This is expected since the bilingual program is largely designed to benefit this sub-group of the population, which usually comprises indigenous individuals in a state of precarity. The large benefits reported in Table 9, however, may accrue through the greater efficiency in teaching (Patrinos & Velez, 2009) and from increased motivation and self-esteem arising from the new opportunities presented to rural students (Lesniewski, 2015).

Table 9: Regression Results – Rural Individuals

Variable Years Schooling Literacy Numeracy Primary School Completion Interaction 0.754*** 0.366*** 0.161*** 0.122*** 0.185*** 0.151*** 0.067 0.034 (0.093) (0.083) (0.054) (0.026) (0.046) (0.019) (0.049) (0.022) Treatment 2.60*** 0.853*** 0.237*** 0.208*** 0.224*** 0.232*** 0.243*** -0.402*** (0.352) (0.078) (0.025) (0.014) (0.025) (0.013) (0.033) (0.045) Exposed 0.006 0.104 0.206*** 0.115*** 0.287*** 0.129** 0.049 0.074 (0.594) (0.371) (0.046) (0.037) (0.106) (0.038) (0.049) (0.049)

XitsIncluded No Yes No Yes No Yes No Yes

(24)

23

Finally, Table 10 evidences that indigenous individuals have too increased their education outcomes. Literacy, numeracy and primary schooling rose by 16.4, 17.2 and 17.1 percent, respectively. These figures are significant at the 1 percent level and robust to the inclusion of control variables. Years of schooling also experienced increases, albeit the impact being insignificant. Again, these results are largely expected, not least because the reform was designed with the indigenous populations of Bolivia in mind. Psacharopoulos (1993) and Hynsjo & Damon (2016) amongst others, corroborate these positive impacts and hold that the use of familiar languages is of great benefit to these subgroups. As a result, Hypothesis 3, which claimed traditionally disadvantaged groups would benefit from the reform, is confirmed.

Table 10: Regression Results – Indigenous Individuals

Variable Years Schooling Literacy Numeracy Primary School Completion Interaction 0.635 0.239 0.164*** 0.066*** 0.172*** 0.094*** 0.171*** 0.162*** (0.574) (0.266) (0.026) (0.011) (0.026) (0.014) (0.028) (0.019) Treatment 2.409*** 1.644*** 0.211*** 0.087*** 0.211*** 0.098** 0.215*** 0.185*** (0.340) (0.268) (0.020) (0.009) (0.019) (0.011) (0.028) (0.020) Exposed 0.755 1.025** 0.080** 0.093*** 0.094** 0.106*** 0.069 0.074* (0.574) (0.301) (0.026) (0.019) (0.026) (0.027) (0.051) (0.030)

VIII. Conclusion

The results indicate that bilingual education in Bolivia has been successful at increasing education outcomes. In particular, the investigation found that years of schooling, literacy, numeracy and primary school completion increased by 0.514 years, 8.6, 8.7 and 10.9 percent, respectively. In addition, the results found for female and rural individuals largely second these improvements. Females saw years of schooling, literacy, numeracy and primary school completion increase by 0.722 years, 9.1, 6.2 and 11.7 points respectively. Rural individuals experienced increases in years of schooling, literacy, numeracy and primary school completion of 0.754 years, 16.1, 18.8 and 6.7 percent. Similarly, indigenous individuals saw an increase of 16.4, 17.2 and 17.1 percent for literacy, numeracy and primary schooling, respectively. These results are largely significant at the 1 percent level, with the exception of rural individuals’ primary school completion. Furthermore, the results are robust to the

(25)

24

inclusion of control variables and the redefinition and dropping of Partially Exposed individuals, which served as falsification tests. Therefore, the investigation’s Hypotheses, which predicted that the impact of bilingual education in Bolivia would be positive on education outcomes including for marginalized groups, have been confirmed.

These results largely corroborate those found in the pedagogical literature and empirical evidence. The improvements in education are likely the result of enhanced teaching efficiency that accelerates learning (Hynsjo & Damon, 2016), while the increases in school completion presumably arise from increased demand for education from both parents and students (Enge & Chesterfield, 1996). Similar explanations may be found for the increases in education for marginalized groups. Enhanced teaching efficiency that improves learning and increased demand for education have also been found for females and rural individuals across the literature. However, the investigation could have increased its validity had it used data on academic achievement to truly assess the extent of learning in Bolivia. This was not possible due to a lack of test score data from Bolivia; yet, other investigations would largely benefit from the utilization of such data to evaluate the impact of bilingual education.

Nevertheless, the results would indicate that expanding bilingual schooling across the developing world, particularly in countries that are home to large indigenous populations and where instruction occurs in a non-native language would be hugely beneficial. In addition, the investigation reveals that not only small-scale programs that are confined to a small subset of schools or geographical areas are advantageous. ASEP, which is the largest-scale bilingual education program to knowledge, has been quite successful at improving the education situation of a country in dire need and affected by great scarcity. The challenge for future large-scale programs in developing countries would involve the allocation of resources, where these are typically scarce. However, evidence has also revealed that the cost savings and efficiency gains from bilingual education in developing countries are vast. Patrinos & Velez (2009) have found how a bilingual schooling program in Guatemala was able to generate cost savings equal to the cost of primary education for 100,000 students from reduced repetition alone. It is thus very plausible that the increased scale of nation-wide programs would result in similar if not larger cost savings from economies of scale. Future research could thus advocate for and investigate other large-scale bilingual programs in developing country settings, making emphasis on the extent of their efficiency gains and cost savings. Revealing the true welfare gains are essential for the successful implementation and expansion of programs such as bilingualism, which aim to increase the quality and efficiency of education. These have the potential to radically change the education landscape across the developing world, which is imperative to remedy the prolonged dismal situation that continues to devastate those in dire need.

(26)

25 Works Cited

Andersen, L. E. (2002). Rural-urban migration in Bolivia: Advantages and disadvantages (No. 05/02). Documento de Trabajo, Instituto de Investigaciones Socioeconómicas, Universidad Católica Boliviana.

Awopetu, A. V. (2016). Impact of Mother Tongue on Children's Learning Abilities in Early Childhood Classroom. Procedia-Social and Behavioral Sciences, 233, 58-63.

Baker, C. (2011). Foundations of bilingual education and bilingualism (Vol. 79). Multilingual matters. Bebbington, A., Abdulai, A. G., Bebbington, D. H., Hinfelaar, M., & Sanborn, C. (2018). Governing

Extractive Industries: Politics, Histories, Ideas. Oxford University Press.

Benson, C. J. (2000). The primary bilingual education experiment in Mozambique, 1993 to 1997. International Journal of Bilingual Education and Bilingualism, 3(3), 149-166.

Benson, C. (2005). Girls, Educational Equity and Mother Tongue-Based Teaching. UNESCO Bangkok. Asia and Pacific Regional Bureau for Education, PO Box 967, Prakhanong Post Office, Bangkok 10110, Thailand.

Cardozo, M. T. L. (2012). Decolonising Bolivian education: Ideology versus reality. In Logics of Socialist Education (pp. 21-35). Springer, Dordrecht.

Cardozo, M. L., Strauss, J. A., & Muhr, T. (2013). From the Local to the Regional and Back: Bolivia’s Politics of Decolonizing Education in the Context of the ALBA--TCP. In Counter-Globalization and Socialism in the 21st Century: The Bolivarian Alliance for the Peoples of Our America (pp. 203-218). London Routledge.

Crevels, Mily y Muysken, Pieter (2009). Lenguas de Bolivia: presentación y antecedentes. En: Mily Crevels y Pieter Muysken (eds.) Lenguas de Bolivia, tomo I Ámbito andino, 13-26. La Paz: Plural editores.

Curiel, H., Rosenthal, J. A., & Richek, H. G. (1986). Impacts of bilingual education on secondary school grades, attendance, retentions and drop-out. Hispanic Journal of Behavioral Sciences, 8(4), 357-367.

Dalby, D. (1985). The educational use of African languages in sub-Saharan Africa: The state of the art. In Document ED/85/WS/62. Unesco Paris.

Dutcher, N., & Tucker, G. R. (1996). The use of first and second languages in education-a review of international experience. The World Bank.

d'Emilio, L. (2001). Voices and processes toward pluralism: Indigenous education in Bolivia. Swedish International Development Cooperation Agency (Sida).

Enge, K. I., & Chesterfield, R. (1996). Bilingual education and student performance in Guatemala. International Journal of Educational Development, 16 (3), 291-302.

(27)

26

Fuller, B., & Clarke, P. (1994). Raising school effects while ignoring culture? Local conditions and the influence of classroom tools, rules, and pedagogy. Review of educational research, 64(1), 119-157.

Garcia de Rosa, S., Guadalupe, C., & Ruiz Pozuelo, J. (2015). The Latin America Learning Barometer: Moving forward in access, lagging behind in learning. Global Economy & Development, Brookings Institute.

Glewwe, P. (2002). Schools and skills in developing countries: Education policies and socioeconomic outcomes. Journal of economic literature, 40(2), 436-482.

Hanushek, E. A. (1995). Interpreting recent research on schooling in developing countries. The World Bank Research Observer, 10 (2), 227-246.

Hornberger, N. H., & Chick, J. K. (2001). Co-Constructing School. Voices of Authority: Education and Linguistic Difference, 1, 31.

Howard, R. (2009). Education reform, indigenous politics, and decolonisation in the Bolivia of Evo Morales. International Journal of Educational Development, 29(6), 583-593.

Hynsjö, D., & Damon, A. (2016). Bilingual education in Peru: Evidence on how Quechua-medium education affects indigenous children's academic achievement. Economics of Education Review, 53, 116-132.

INE. Bolivia Population and Household Census. (2012). Instituto Nacional de Estadistica de Bolivia. INE. Bolivia Population and Household Census. (2017). Instituto Nacional de Estadistica de Bolivia. Infante, I., Letelier, M. E., & Rivero, J. (2013). Nunca es tarde para volver a la escuela: resultados de

la encuesta del UIS sobre programas de educación de adultos y alfabetización en América Latina y el Caribe. UNESCO. Instituto de Estadística.

Jimenez-Quispe, Luz. (2011). “We Want Public Education! he New”. Bolivia: Revolutions and Beyond. ReVista: Harvard Review of Latin America. Cambridge.

Lesniewski, N. (2015). Educational Attainment in Guatemala: An Analysis of Characteristics Related to Youth Perspectives and Households. MPA/MPP Capstone Projects. 232

McEwan, P. J. (2008). Can schools reduce the indigenous test score gap? Evidence from Chile. The journal of development studies, 44(10), 1506-1530.

Patrinos, H. A., & Velez, E. (2009). Costs and benefits of bilingual education in Guatemala: A partial analysis. International Journal of Educational Development, 29(6), 594-598.

Postero, N. (2010). Morales’s MAS government: building indigenous popular hegemony in Bolivia. Latin American Perspectives, 37(3), 18-34.

Prevatt, F., & Kelly, F. D. (2003). Dropping out of school: A review of intervention programs. Journal of school psychology, 41(5), 377-395.

(28)

27

Psacharopoulos, G. (1993). Ethnicity, education, and earnings in Bolivia and Guatemala. Comparative Education Review, 37(1), 9-20.

Punch, S. (2004). The impact of primary education on school-to-work transitions for young people in rural Bolivia. Youth & Society, 36(2), 163-182.

Seid, Y. (2016). Does learning in mother tongue matter? Evidence from a natural experiment in Ethiopia. Economics of Education Review, 55, 21-38.

Sichra, I. (1992). Informe final de evaluación. Test de salida primer grado. Área Andina del Proyecto de Educación Intercultural Bilingüe (PEIB). Ministerio de Educación, Bolivia.

Storm, Helen. (2011). “The New Bolivian Education Law”. Bolivia: Revolutions and Beyond.ReVista: Harvard Review of Latin America. Cambridge.

Talavera Simoni, M. L. (2013). La educación pública frente a una nueva reforma: Avances de investigación. Estudios bolivianos, 35.

UNESCO. (2018). Out of School Children and Youth. Unesco.

Walter, S. L., & Dekker, D. E. (2011). Mother tongue instruction in Lubuagan: A case study from the Philippines. International Review of Education, 57(5-6), 667-683.

Widmark, C. (2002). Education as Expanding Freedoms in Bolivia. Education—A Way out of Poverty, 79-95. Swedish International Cooperation Agency.

World Bank. (2017). World development report 2018: Learning to realize education’s promise. The World Bank Group, 2018.

To What Extent has Bilingual Education Increased Education Outcomes in Bolivia?