Disparity in school children's reading skills in 11 African countries

Zhang, Huafeng; Holden, Stein Terje Working Paper Disparity in school children's reading skills in 11 African countries Centre for Land Tenure Studies Working Paper, No. 05/24 Provided in Cooperation with: Centre for Land Tenure Studies (CLTS), Norwegian University of Life Sciences (NMBU) Suggested Citation: Zhang, Huafeng; Holden, Stein Terje (2024) : Disparity in school children's reading skills in 11 African countries, Centre for Land Tenure Studies Working Paper, No. 05/24, ISBN 978-82-7490-327-2, Norwegian University of Life Sciences (NMBU), Centre for Land Tenure Studies (CLTS), Ås, https://hdl.handle.net/11250/3158906 This Version is available at: https://hdl.handle.net/10419/306772 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Centre for Land Tenure Studies Working Paper 05/24 ISBN: 978-82-7490-327-2 Norwegian University of Life Sciences (NMBU) Disparity in School Children's Reading Skills in 11 African Countries Huafeng Zhang and Stein T. Holden 1 Disparity in School Children's Reading Skills in 11 African Countries Huafeng Zhanga,b* and Stein T. Holdena a School of Economics and Business, Norwegian University of Life Sciences, P.O. Box 5003, 1432 Ås, Norway b Fafo Institute for Labour and Social Research, Borggata 2B, Postboks 2947, Tøyen. 0608 Oslo, Norway * Corresponding author. Email: [email protected] Abstract To promote SDG Goal 4 and "education for all", this study investigates children’s basic reading skills in 11 low-income and lower-middle-income African countries, using standardized reading tests from the Multiple Indicator Cluster Surveys (MICS). Research specifically examining children’s reading skills and disparities across socioeconomic groups in African contexts remains scarce. This study addresses a critical knowledge gap by providing comparative evidence on reading skills disparities across diverse social backgrounds, including children with disabilities. Our study provides new evidence on the “Learning Crisis in the Global South”, revealing alarmingly low levels of reading skills but with considerable variation across the 11 African countries studied. Substantial reading skills differences exist between children from disadvantaged backgrounds—those with disabilities, living in rural areas, and from poorer, less educated families—and their non-disadvantaged peers. Notably, these disparities are often more pronounced in countries with higher overall reading proficiency. Moreover, there are persistent gaps between children with and without disabilities across the countries and socioeconomic groups in this study. Encouragingly, children with disabilities benefit from improved socioeconomic conditions just as much as non-disabled children. These findings underscore the diverse challenges faced by children from different disadvantaged backgrounds in varying contexts. Keywords: Africa, Children with disabilities (CWD), Educational inequality, Poverty, Reading skills, Socioeconomic background, Urban-rural disparity JEL codes: I24: Education and Inequality 2 1. Introduction The UN Sustainable Development Goal 4 underscores the importance of achieving inclusive and equitable quality education for all (UN, 2015; UNESCO, 2016). There is a growing interest in understanding the educational outcomes of children from disadvantaged backgrounds and identifying the factors that contribute to variations in these outcomes, which can inform the development of effective educational policies (Evans & Mendez Acosta, 2021; Bashir et al., 2018; Musau, 2018). In recent years, following the debate on the “Learning Crisis in the Global South” (World Bank, 2018; UNESCO, 2014), reading proficiency has emerged as a crucial focus in sub-Saharan Africa, recognized as a key indicator of learning outcomes and the success of formal education. The percentage of students attaining the minimum proficiency level in reading skills is a key indicator for achieving SDG Goal 4, given the emphasis on reading skills by the UNESCO Global Education Monitoring Report (2014). Previous research in developed contexts has emphasized the persistent differences in reading skills between children from disadvantaged and non-disadvantaged backgrounds (Hernandez, 2011; Heckman, Pinto & Savelyev, 2013; Dolean et al., 2019). In developing countries, efforts have traditionally centred on socioeconomic factors such as gender, education, income, and geographical location (Zhang, 2006; Clercq, 2020; Chmielewski, 2019). Numerous cross-country studies on children’s reading performance have offered valuable insights into the role of gender, home environment, school socioeconomic status, and literacy interventions in shaping children’s reading (León et al., 2022; Kim et al., 2020; Park 2008; Chiu and McBrideChang 2006, 2010; Shiel and Eivers 2009). However, these studies often rely on international standard learning assessments, such as PIRLS (the Progress in International Reading Literacy Study) and PISA (Programme for International Student Assessment). These assessments primarily target developed or OECD countries, with limited participation from African nations. 3 Of the 102 countries that have ever participated in PISA, only eight are from Africa, including just four from Sub-Saharan Africa. PIRLS has even fewer African participants. Due to data constraints, comparative studies on educational outcomes in African countries tend to focus primarily on school enrolment, attendance, and completion rate (Wodon et al., 2018). Research specifically examining children’s learning performance, such as reading or numeracy skills, and the disparities in these outcomes across socioeconomic groups in African contexts, remain scarce. One notable exception is Zhang and Holden (2023), who analysed children’s numeracy skills across eight African countries using MICS data, with a special focus on children with disabilities. The challenges faced by children with disabilities (CWD) and their low learning performance have only recently garnered attention, particularly following the adoption of the United Nations Convention on the Rights of Persons with Disabilities (UNCRPD) in 2006 (UN, 2006). Recent studies have made efforts to understand the schooling challenges faced by CWD, focusing on differences in school access, attendance and enrolment in developing countries (Filmer, 2008; Mizunoya et al., 2018; UNESCO, 2018). However, studies specifically addressing how much CWDs are falling behind in reading skills learning are rare in the context of developing countries, with only a few exceptions from individual studies in Asia (Singal et al., 2020), and none in the African context. Based on nationally representative data across 11 low-income and lower-middle-income African countries, we evaluate the reading skills of children aged 10 to 14 years old and investigate variations in reading skills across rural versus urban areas, between children with disabilities (CWD) versus children without disabilities (CWOD), as well as between children from poor and less educated families versus better-off and more educated families. More specifically, we assess the relative performance of CWD vs. CWOD within various social groups as well as examine how these disparities vary across different African countries. 4 Our research aims to answer the following research questions: 1) To what extent do children from disadvantaged backgrounds (e.g., children from poor or less educated families, rural areas, or with disabilities) lag behind their peers (children from more affluent or educated families, urban areas, or without disabilities) in basic reading skills? 2) Do disadvantaged children benefit equally from improvements in their country’s overall reading proficiency? 3) Can improvements in micro-level social factors help mitigate the learning constraints faced by children with disabilities? This paper is unique in its exclusive focus on school children’s reading skills performance across low-income and lower-middle-income African countries, all of which were included in the sixth round of Multiple Indicator Cluster Surveys (MICS) between 2017 and 2020. First, we present comprehensive, nationally representative evidence of the substantial variation in basic reading skills among children from different socioeconomic backgrounds. We employ consistent, standardized tests and measurements of reading skills both within and across countries. We identify substantial differences in reading skills across the 11 countries, as well as across socioeconomic groups within each country. Second, we utilize the standardized identification of children with disabilities in the MICS survey to assess their reading skills, using children without disabilities in each country as a counterfactual. Overall, children with disabilities lag behind children without disabilities. However, an interesting finding is that children with disabilities in better-performing countries outperform children without disabilities in other countries. This suggests that children with disabilities benefit from strong educational systems as much as children without disabilities in terms of improving their basic reading skills. 2. Conceptual framework Reading skills are crucial for the development of various other academic skills in school and can greatly impact children's likelihood of repeating grades or dropping out (Reschly, 2010). 5 Several social, familial and individual factors influence children's learning, and the mechanisms through which these factors influence learning are multifaceted (Taylor & Yu, 2009). Pace et al. (2017) identify three potential pathways by which socioeconomic status might influence children’s language development, which are child characteristics, parent-child interaction, and the availability of learning resources. This paper aims to evaluate children’s reading skills performance among children who are disadvantaged in any of the three potential pathways as suggested by Pace et al. (2017). First, children who have functional challenges in one of the four main functional domains – vision, hearing, physical, intellectual – or with multiple functional challenges. Second, children from poor families, defined as those in the lowest quintile of the asset index, and children from families without schooling. These children quite often have little access to critical learning resources and parental engagement for language development. Finally, children living in rural areas, where learning resources are constrained and school quality is often lower. Families with higher social status, including better income and higher education levels, tend to provide better support for their children's learning. Children from more advantaged backgrounds often begin their learning process earlier than their peers from disadvantaged families (Lee & Burkham, 2002). Additionally, they may indirectly benefit from residing in neighbourhoods with higher-quality schools (Anderson, Case and Lam, 2001). Parents with higher social status are also more likely to actively engage with the school community, thereby contributing to overall school quality. The neighbourhood environment can influence children's learning outcomes. In the African context, although not extensively studied, there is evidence of urban-rural disparities in schooling (Zhang, 2006). Rural areas often face challenges related to school quality due to a lack of infrastructure, educational resources, and qualified teachers. Furthermore, in neighbourhoods characterized by high levels of poverty in rural areas, various social issues 6 affecting disadvantaged families can be exacerbated. Children are also exposed to the influences of their peers in the same neighbourhood or school (Leventhal and Brooks-Gunn, 2001). The challenges related to learning reading skills vary greatly across different disability types due to the diverse nature of functional difficulties (Premeaux, 2001; Anastasiou & Kauffman, 2011). Children with vision disabilities may have the same capability to develop reading skills as their peers, but the real challenges often stem from the availability of aids, such as corrective lenses, optical devices, and glasses (Le Fanu et al., 2022), as well as access to consultative instructional services (Corn & Koenig, 2002). For children with hearing disabilities, the challenge of learning to read often arises from a lack of exposure to their first language before the critical period (Kushalnagar et al., 2010). This puts them at high risk of linguistic deprivation (Mayberry, 1994). Children with physical disabilities may not face apparent functional challenges in learning reading skills, but they frequently experience high rates of school absenteeism due to factors like long distances to school and lack of infrastructure, materials, and support (Tanya et al., 2023). Children with intellectual disabilities struggle with developing reading skills due to challenges in various abilities, including information processing, cognitive abilities, and attentive behaviours (Tolar et al., 2016; Chan & Dally, 2001). Children with multiple disabilities are exposed to higher risks related to several different functional challenges. Moreover, the availability of appropriate teaching materials and pedagogical interventions for CWD can enhance their skill development. We set up the first hypothesis concerning the role of factors related to child characteristics, parent-child interaction, and the availability of learning resources: H1. The percentages of school children aged 10-14 with satisfactory reading skills among children with a) families in the lowest quintile of asset index, b) families without schooling, c) rural residence, d) disabilities (vision, hearing, physical, intellectual, and multiple 7 disabilities) are significantly lower than that among other children without such disadvantaged backgrounds. Several cross-country studies, focusing on school enrolment, have shown that disparities in enrollment and attendance for disadvantaged children are more pronounced in countries with higher overall enrollment rates and better socio-economic development (Filmer, 2008; Mizunoya et al., 2018; Lewis et al., 2022). We formulate the second hypothesis to explore whether disadvantaged children benefit equally from improvements in their country’s overall reading proficiency: H2. The differences in the percentage of school children with satisfactory basic reading skills are more pronounced in countries with higher overall reading proficiency, when comparing a) poor (children from families in the lowest quintile of asset index) vs. non-poor, b) children from families without vs. with schooling, c) rural vs. urban children, and d) CWD vs. CWOD. The fundamental question revolves around whether CWD, when raised in families with a more advantageous social background (urban residence, higher income, higher education), can successfully bridge the academic performance gap compared to CWOD. Can improvements in micro-level social factors help mitigate the learning constraints faced by children with disabilities? We set up the third hypothesis related to the reading skills associated with children's disabilities across different social groups: H3. The differences in the percentage of school children with satisfactory basic reading skills between CWD and CWOD are smaller in a) urban, b) higher-income, c) more educated families. Our H3a-c hypotheses are based on the notion that families with advantageous conditions can better support CWD in overcoming learning challenges. Finally, due to data 14 Based on the overall reading skills proficiency of these countries, we can categorize them into three groups: low-reading countries, which include the Central Africa Republic, Chad, DRCongo, and The Gambia; mid-reading countries, which include Ghana, Madagascar, Malawi, and Togo; and high-reading countries, which include Lesotho, Tunisia, and Zimbabwe. 4.2 Reading skills across micro-level factors In the first set of regressions, we run inverse probability weighted7 pooled least squares regression models by including one of the four micro factors in each of four models: 1) household asset index quintile, 2) family members’ highest educational level, 3) location (rural vs. urban), and 4) disability status. The final regression, labelled as Model 5, includes all the micro-level factor variables and control variables (Table 4). Table 4 Here Table 4 indicates large differences in the share of school children with satisfactory reading skills across various groups. Children from the wealthiest quintile of the asset index outperform those from the poorest quintile by 37 percentage points (Model 1). Children in families with primary education show a 6 percentage-point advantage over those from families without any schooling, while those from families with a member has completed junior secondary education or higher achieve 21 percentage-point advantage (Model 2). In the full model incorporating all factors, the coefficients for wealth and education from Models 1 and 2 are reduced, likely reflecting a correlation between these factors. Urban children outperform their rural counterparts by 23 percentage points in satisfactory reading skills before accounting for micro-level factors (Model 3) and by 9 percentage points after these factors are controlled for (Model 5). 7 The outputs for the first stage of selection model are presented in Appendix I. 15 Compared to CWOD, children with hearing disabilities (15 percentage points lower), intellectual disability (16 percentage points lower) and multiple disabilities (17 percentage points lower) exhibit lower proficiency rates (Model 4). The finding remains consistent with or without controlling for other factors (Model 4 and 5). 4.3 Disparities in reading skills across 11 African countries To test hypothesis H2, we include country-specific dummy variables and interaction terms between micro-level factors and individual countries. Figure 1 presents the estimated proportion of 14-year-old children with satisfactory reading skills across various disadvantaged groups: rural children, children with disabilities (CWD), children from poor families in the lowest quintile of the asset index, and children from families without schooling. The figure also includes data on children who do not belong to these disadvantaged groups, offering a comparative analysis across the 11 African countries in our sample. Figure 1 Here Disparities in reading skills between children from poor and non-poor families are significantly larger in countries with mid-level reading proficiency, such as Ghana (23 percentage points), Madagascar (23 percentage points), Togo (15 percentage points), and Zimbabwe (14 percentage points), and Lesotho (10 percentage points). In contrast, these disparities are much smaller in countries with low reading proficiencies, such as Chad (8 percentage points) and DRCongo (6 percentage points), or even no significant disparities, such as in the Central Africa Republic and The Gambia. In Tunisia, where most children have high basic reading proficiency, the differences are also insignificant. An exception is Malawi, which, despite having mid-level reading proficiency, shows no significant disparity between poor and non-poor children. Disparities in reading skills between children from families with and without 16 schooling have largely mirrored those between poor and non-poor children, with much lower disparities in countries with overall low reading proficiency. Urban-rural disparities in reading skills are the most pronounced in Ghana (24 percentage points), Togo (22 percentage points) and Zimbabwe (21 percentage points), while they are significant but small in DRCongo (12 percentage points), Lesotho (12 percentage points), and Madagascar (8 percentage points). For other countries, the urban-rural disparities are not significant. Disparities in reading skills for children with disabilities (CWD) are significant across all 11 African countries, ranging from 7 to 22 percentage points. The largest disparity is observed in the Gambia, while countries with lower reading proficiency show smaller differences. The sample size for CWD is quite limited in several countries, resulting in a large variance in the estimated outcomes for CWD. Consequently, we further analyze the data across the three country groups defined in Section 4.1 (Fig 2). The results from group-level analysis are similar to those from the country-level analysis. Disparities in reading skills between children from poor and non-poor families and between children from families with and without schooling are not significant in low-reading countries but are much larger in mid-reading and high-reading countries. The urban-rural disparity is especially high in the high-reading countries. However, disparities between CWD and CWOD remain consistently significant across countries with different levels of reading proficiency. Fig 2 Here 4.4 Disparities in reading skills related to disabilities To test hypothesis H3, we include all micro-level indicators, as well as the interaction terms between disability status and other micro-level indicators (urban/rural residence, wealth index, 17 and family's highest educational level) in the country fixed effect model. The regression results at various cutoff points are presented in Appendix II. Figure 3 displays the estimated proportion of 14-year-old children with satisfactory reading skills. These predictions are made with covariates set at their means for both CWD and CWOD in different social groups (urban vs. rural, high vs. low socio-economic status, more vs. less educated families). These disparities in reading skills between CWD and CWOD in schools are visually represented as lines connecting two estimated reading skill proficiency rates in various social groups. A steeper incline in the line indicates a higher disparity between CWD and CWOD, while a flatter line suggests a smaller disparity. Figure 3 Here Figure 3 suggests that disparities in reading skills proficiency between CWD and CWOD do not vary significantly across different social groups. These disparities remain relatively constant at around 15 percentage points in various groups. The most significant disparities are observed in urban areas (19 percentage points) and among families without any schooling (21 percentage points). Furthermore, it is noteworthy that CWD in social groups with advantaged backgrounds (urban, rich and more-educated families) have achieved similar levels of reading skill proficiency as their CWOD peers in social groups with disadvantaged backgrounds (rural, economically disadvantaged, and less-educated families). 5. Discussion and study limitations 5.1 Discussion In this section, we will discuss the findings related to the key hypotheses. We will also discuss important limitations of our study and provide some suggestions for future research. 18 Utilizing a standardized reading test, the paper reveals particularly low overall reading skills and considerable variations among school children across the 11 African countries. The proportion of school children attaining satisfactory reading skills ranges widely, from 18 percent in the Central Africa Republic to 88 percent in Tunisia. In our combined sample from these 11 countries, less than half (45 percent) of the school children have reached a satisfactory reading level, namely, they are able to read the basic text properly. It is important to note that there is substantial variation in the level of school attendance across these countries, with rates ranging from 43 percent in Chad to 69 percent in Madagascar, and reaching as high as 95 percent in Lesotho, Malawi, and Tunisia. Since we expect a much lower reading skill level for children not enrolled in school, the overall reading skill level and actual gap in learning across these countries is likely higher when differences in school attendance are considered. For instance, while the average reading skill proficiency rate is 21 percent among schoolchildren in Chad, school attendance is only 43 percent. The first set of models support hypothesis H1, showing that children from 1a) impoverished backgrounds, 1b) less-educated households, and 1c) rural areas, exhibit significantly lower reading skills than their peers from affluent families, more educated households, or urban areas. Hypothesis H1d) is only partially supported: the percentage of school children with satisfactory reading skills is significantly lower among those with hearing, intellectual, and multiple disabilities8 compared to their CWOD peers. However, it is important to note that children with vision or physical disabilities do not significantly lag behind, and the conclusion regarding children with hearing disabilities does not remain statistically significant when all control variables are included in the analysis. 8 The coefficient for children with multiple disabilities become insignificant when more control variables included. The sample size is quite limited due to the very low school attendance in this group of children, which may lead to high standard error. 19 As demonstrated by numerous studies in developed contexts (Pace, etc., 2017), children from disadvantaged backgrounds tend to lag behind in reading abilities. Notably, our analysis shows that family poverty has the strongest correlation with children's reading skills. The proportion of school children in the richest quintile group who have achieved satisfactory reading skills is approximately 24-35 percentage points higher than those in the poorest quintile group. What is particularly notable in our study is the observation that a substantial proportion of school children obtain extreme values in their reading test scores, either very low or very high scores. The concern here is primarily for school children who, at their current age, continue to achieve very low scores in basic reading tests. This underscores the substantial challenges they may have encountered in developing proficient reading skills in the long future. Children from disadvantaged backgrounds are particularly representative. Furthermore, our study indicates that school children with vision and physical disabilities do not exhibit significant differences in their reading skills compared to the nondisabled children. It is plausible that they have managed adequately with basic reading skills. However, if more comprehensive reading tests were to be introduced, these children might also encounter challenges and potential difficulties in meeting advanced reading skill requirements. Our findings support Hypothesis H2a, H2b, and H2c, indicating that disparities in reading proficiency rates across socioeconomic groups and urban-rural disparities are more pronounced in countries with higher overall reading proficiency. In countries with very low reading proficiency, such as the Central African Republic (average reading skills score of 18 percent), Chad (21per cent), and DRCongo (19 percent), disparities in reading skills across socioeconomic groups are either insignificant or much smaller compared to other countries. The largest disparities across socioeconomic groups are observed in countries with midlevel reading proficiency, such as Ghana (47 percent), Madagascar (51 percent), Togo (38 20 percent), and Zimbabwe (56 percent). Urban-rural disparities are also most pronounced in countries with relatively high reading proficiency. However, in Tunisia, which boasts the highest level of socio-economic development and the highest reading proficiency (88 percent) among the 11 countries, no significant disparities in reading skills are found among children from different disadvantaged backgrounds. Our findings do not support Hypothesis H2d, which posits that disparities in reading proficiency rates between children with and without disabilities are more pronounced in countries with higher overall reading proficiency. Meanwhile, Tunisia, the country with the highest reading proficiency (88 percent), exhibits relatively high disparities in reading skills between CWD and CWOD. However, a closer examination shows that the gap of 20 percentage points, when considered in proportion to the overall proficiency level, is not larger compared to the 7-12 percentage points gaps observed in countries with significantly lower reading proficiency, such as the Central African Republic, Chad, and DRCongo (with overall proficiency levels ranging from 18 to 21 percent). In countries with mid-level reading proficiency (35-58 percent), disparities between CWD and CWOD range from 12 to 25 percentage points, further suggesting that disability-related disparities are not significantly different across countries with different reading proficiency. Our findings do not support Hypothesis H3 that disparities in the percentage of school children with satisfactory reading skills between CWD and CWOD would be less pronounced in households with more advantaged backgrounds. Instead, these disparities have remained relatively constant across different social groups. It is worth emphasizing that these results are based on children who are currently enrolled in school. When we consider out-of-school children, recognising the overrepresentation of CWD in this group, it becomes apparent that disparities in social groups with disadvantaged backgrounds may have been underestimated. 21 However, as long as children are enrolled in school, a consistent gap between CWD and CWOD appears to persist. 5.2 Study limitations Several limitations should be considered when interpreting the findings of this study. First, the reading test used in the MICS survey is relatively basic. Given the age range of children tested (10-14 years), it may not comprehensively assess more advanced reading skills. However, even with the basic test, the prevalence of satisfactory reading skills among children aged 10-14 in most of these countries is notably low, indicating limited reading abilities across many African countries. Introducing a more comprehensive reading test could potentially reveal even greater difficulties, especially among children from disadvantaged backgrounds, and is likely to expose even larger disparities in reading proficiency. Second, it is crucial to recognize that this study exclusively focuses on children currently enrolled in school. Many children not attending school and therefore not taking the reading test are disproportionately from disadvantaged backgrounds. As a result, the disparities estimated in this group may have been underestimated. Moreover, there is substantial variation in school attendance rates across the countries studied. Careful consideration is needed when analysing countries with low school enrolment. It is important to emphasize that the conclusions drawn in this paper are applicable exclusively to children enrolled in school and cannot be generalized to encompass all children in these countries. Third, the selection of countries in this study was not guided by strict predefined criteria but was rather constrained by data availability. It is essential to interpret the estimated disparities cautiously due to the inherent randomness associated with the selection of countries in this paper. 22 6. Conclusion Our study provides new evidence on the reading proficiency of school children aged 10-14 across 11 African countries, drawing from unique nationally representative data. Through a standardized reading test, the paper uncovers notably low overall reading skills and significant disparities among school children across 11 African countries. By examining the correlations between diverse regional, familial, and individual factors, we aimed to uncover important factors that may influence school children's acquired reading skills. Benefiting from the large sample size from country-pooled data in the MICS standardized data, this study emphasizes the heterogeneous disability effect on children’s reading skills related to disability type, which has been overlooked by many studies due to sample size limits. A comparative analysis across 11 African countries suggests that disparities in reading skills among children from disadvantaged backgrounds are non-existent or minimal in countries with low overall reading proficiency. In contrast, these disparities are more pronounced in some countries with mid-level reading proficiency. Notably, despite having the highest overall reading proficiency, Tunisia shows no significant differences in reading skills across the social groups examined. On the other hand, given the basic nature of the reading test in this study, we can only conclude that there are no significant disparities in basic reading skills among disadvantaged children in Tunisia. However, larger disparities may emerge if more comprehensive reading skills are assessed. Another unique contribution of our study lies in its findings related to children with disabilities (CWD), a topic that has received relatively little attention in recent literature, likely due to data limitations. Our study highlights a persistent gap in reading skills between CWD and CWOD across countries and various social groups, underscoring the unique challenges CWD faces. Interestingly, improvements in micro-level conditions have not impacted these 23 gaps. Nonetheless, it is encouraging to note that the proportion of CWD with adequate reading skills increases similarly to CWOD in response to improved conditions. This paper underscores the critical role of micro-level socioeconomic factors in addressing challenges faced by vulnerable populations and enhancing reading skills for all. However, certain vulnerable groups, such as CWD, encounter unique challenges in acquiring reading skills. While CWD can make similar gains to CWOD when school quality and socioeconomic conditions improve, a persistent gap between these groups remains. Further targeted and in-depth research is essential to understand the underlying dynamics and identify tailored interventions, which extend beyond the scope of this paper. 30 Figure 3 Estimated proportion of 14-year-old children with satisfactory reading skills for CWD and CWOD across various social groups (Country FE), with 95% confidence intervals. Note: The predictions are calculated at the means of covariates across all countries, with separate predictions for various social groups related to rural and urban residences, family wealth index, and the highest educational level among household members. 31 Appendix Appendix I Regression results from first stage of selection model for each country Variable Central Africa R. Chad DRCongo Ghana Lesotho Madagasc ar Malawi The Gambia Togo Tunisia Zimbab we Disabled -0.292* -0.07 -0.592*** -0.440*** -0.389* -0.16 -0.309*** -0.714** 0.045 -0.305 0.078 Location (base category: urban) -0.280** -0.253** -0.197** -0.290*** -0.169 0.038 -0.130* -0.212 -0.24 0.189 0.378 Wealth index (base category=Poorest) Second quintile 0.12 0.239* 0.033 0.014 0.251* 0.268** 0.156** -0.184 0.09 -0.166 0.155 Middle 0.102 0.354** 0.253*** 0.208 0.353** 0.367*** 0.272*** -0.231 0.068 0.169 0.167 Fourth quintile 0.300* 0.323** 0.490*** 0.304* 0.282* 0.455*** 0.383*** -0.28 0.117 0.288 0.747** Richest 0.343* 0.507*** 0.811*** 0.256 0.573*** 0.334** 0.539*** 0.096 0.14 0.233 0.664* Highest Educational level in the household (base category=No school) Primary -0.08 0.007 -0.055 -0.104 0.227* 0.105 0.155** 0.148 0.061 -0.109 0.244 Junior High 0.066 0.138 0.066 0.023 0.074 0.11 0.366*** 0.159 -0.094 -0.128 0.28 Senior High+ 0.047 0.085 0.066 0.185 -0.148 0.047 0.568*** -0.038 -0.001 -0.119 0.795 Age (Base category=10) 11 -0.018 0.021 0.053 0.098 0.202 0.12 0.196*** 0.172 0.137 -0.175 0.111 e12 -0.057 0.094 0.166* 0.273** 0.134 0.125 0.303*** 0.254 -0.086 -0.045 0.085 age13 0.056 0.077 0.323*** 0.349*** 0.157 0.261** 0.430*** 0.492** 0.034 -0.167 0.117 age14 0.116 0.203 0.495*** 0.631*** 0.176 0.336*** 0.571*** 0.512*** 0.241 -0.223 0.222 Gender (Base category: Boys) -0.167* 0.063 -0.091 -0.099 0.290*** 0.058 0.258*** 0.222* -0.066 -0.019 0.371** * Constant 1.133*** 0.790*** 0.889*** 1.704*** 0.458 0.29 -0.131 1.068** 1.934*** 1.681*** -0.08 Sample size 1458 1910 3468 3159 1823 2972 6332 1355 1663 1669 2144 Appendix II Sensitivity test to the selection of different cutoff thresholds for the outcome variable of reading proficiency. Regression results for the first hypothesis with cutoff points at 80% and 90% are presented in Table II.1 and Table II.2. Regression results for the second hypothesis with cutoff points at 80%, 85%, and 90% are presented in Table II.3. Regression results for the third hypothesis with cutoff points at 80%, 85%, and 90% are presented in Table II.4. No large sensitivity to the selection of different cutoff thresholds is detected. Table II.1 IPW least squares regressions by micro-level factors (outcome variable cutoff at 80%) Model1 Model2 Model3 Model4 Model5 Wealth index (base category=Poorest) Second quintile 0.057*** 0.042*** (0.010) (0.010) Middle 0.112*** 0.080*** (0.009) (0.010) Fourth quintile 0.208*** 0.146*** (0.010) (0.011) Richest 0.372*** 0.265*** (0.010) (0.012) Highest Educational level in the household (base category=No school) Primary 0.058*** 0.031*** (0.009) (0.009) Junior secondary 0.208*** 0.094*** (0.010) (0.010) Senior secondary or higher 0.211*** 0.082*** (0.011) (0.011) Location (base category: urban) -0.225*** -0.087*** (0.008) (0.009) Disability status (base category: non-disabled) Vision disability 0.032 0.022 (0.036) (0.036) Hearing disability -0.137** -0.096* (0.050) (0.047) Physical disability 0.028 0.064 (0.035) (0.035) Intellectual disability -0.165*** -0.158*** (0.016) (0.016) 32 Multiple disabilities -0.167*** -0.119* (0.050) (0.050) Gender (Base category: Men) 0.037*** 0.042*** 0.040*** 0.043*** 0.035*** (0.006) (0.006) (0.006) (0.006) (0.006) Age (Base category=10) age11 0.064*** 0.067*** 0.065*** 0.073*** 0.063*** (0.009) (0.009) (0.009) (0.009) (0.009) age12 0.117*** 0.121*** 0.116*** 0.120*** 0.115*** (0.009) (0.009) (0.009) (0.009) (0.009) age13 0.170*** 0.176*** 0.171*** 0.175*** 0.170*** (0.009) (0.009) (0.009) (0.009) (0.009) age14 0.216*** 0.227*** 0.222*** 0.229*** 0.215*** (0.009) (0.009) (0.009) (0.009) (0.009) Country (Base category=Central Africa R.) Chad 0.039* 0.077*** 0.082*** 0.033 0.065*** (0.016) (0.018) (0.018) (0.018) (0.017) DRCongo 0.092*** -0.021 0.045** 0.002 0.064*** (0.014) (0.016) (0.015) (0.016) (0.015) Ghana 0.348*** 0.292*** 0.310*** 0.300*** 0.338*** (0.015) (0.017) (0.016) (0.019) (0.015) Lesotho 0.490*** 0.439*** 0.470*** 0.409*** 0.492*** (0.017) (0.017) (0.018) (0.018) (0.017) Madagascar 0.416*** 0.400*** 0.424*** 0.373*** 0.429*** (0.016) (0.017) (0.017) (0.018) (0.016) Malawi 0.378*** 0.370*** 0.426*** 0.334*** 0.407*** (0.014) (0.015) (0.015) (0.015) (0.015) The Gambia 0.239*** 0.227*** 0.173*** 0.164*** 0.238*** (0.018) (0.019) (0.019) (0.020) (0.018) Togo 0.276*** 0.243*** 0.259*** 0.216*** 0.284*** (0.018) (0.019) (0.019) (0.020) (0.017) Tunisia 0.772*** 0.716*** 0.686*** 0.716*** 0.736*** (0.013) (0.014) (0.015) (0.015) (0.014) Zimbabwe 0.440*** 0.369*** 0.440*** 0.385*** 0.431*** (0.015) (0.017) (0.016) (0.018) (0.015) Constant -0.177*** -0.107*** 0.123*** 0.017 -0.116*** (0.016) (0.016) (0.016) (0.015) (0.019) Sample size 23591 23572 23591 23591 23572 R2 0.225 0.186 0.199 0.163 0.237 Table II.2 IPW least squares regressions by micro-level factors (outcome variable cutoff at 90%) Model1 Model2 Model3 Model4 Model5 Wealth index (base category=Poorest) Second quintile 0.055*** 0.042*** (0.009) (0.009) Middle 0.101*** 0.073*** (0.009) (0.009) Fourth quintile 0.186*** 0.131*** (0.009) (0.010) Richest 0.337*** 0.240*** (0.010) (0.012) Highest Educational level in the household (base category=No school) Primary 0.046*** 0.022** (0.008) (0.008) Junior secondary 0.185*** 0.083*** (0.009) (0.009) Senior secondary or higher 0.190*** 0.075*** (0.011) (0.011) Location (base category: urban) -0.203*** -0.078*** (0.008) (0.009) Disability status (base category: non-disabled) Vision disability 0.022 0.013 (0.036) (0.036) Hearing disability -0.127** -0.090* (0.047) (0.045) Physical disability 0.038 0.071 (0.036) (0.036) Intellectual disability -0.148*** -0.141*** (0.015) (0.014) 33 Multiple disabilities -0.142** -0.099* (0.048) (0.048) Gender (Base category: Men) 0.034*** 0.038*** 0.037*** 0.040*** 0.032*** (0.006) (0.006) (0.006) (0.006) (0.006) Age (Base category=10) age11 0.054*** 0.057*** 0.055*** 0.063*** 0.053*** (0.009) (0.009) (0.009) (0.009) (0.009) age12 0.093*** 0.096*** 0.092*** 0.096*** 0.091*** (0.008) (0.009) (0.009) (0.009) (0.008) age13 0.135*** 0.139*** 0.135*** 0.140*** 0.135*** (0.009) (0.009) (0.009) (0.009) (0.009) age14 0.173*** 0.183*** 0.178*** 0.185*** 0.172*** (0.009) (0.009) (0.009) (0.010) (0.009) Country (Base category=Central Africa R.) Chad 0.040** 0.074*** 0.079*** 0.035* 0.062*** (0.015) (0.016) (0.016) (0.016) (0.015) DRCongo 0.073*** -0.030* 0.031* -0.008 0.046*** (0.013) (0.014) (0.013) (0.014) (0.013) Ghana 0.283*** 0.232*** 0.248*** 0.239*** 0.273*** (0.014) (0.016) (0.015) (0.017) (0.014) Lesotho 0.361*** 0.316*** 0.342*** 0.288*** 0.363*** (0.017) (0.018) (0.018) (0.019) (0.017) Madagascar 0.290*** 0.277*** 0.298*** 0.252*** 0.303*** (0.015) (0.016) (0.016) (0.017) (0.015) Malawi 0.258*** 0.251*** 0.301*** 0.218*** 0.284*** (0.013) (0.013) (0.014) (0.014) (0.014) The Gambia 0.196*** 0.183*** 0.137*** 0.128*** 0.193*** (0.016) (0.017) (0.017) (0.018) (0.016) Togo 0.204*** 0.174*** 0.189*** 0.150*** 0.211*** (0.015) (0.016) (0.016) (0.017) (0.015) Tunisia 0.680*** 0.630*** 0.602*** 0.629*** 0.647*** (0.014) (0.015) (0.015) (0.015) (0.015) Zimbabwe 0.428*** 0.365*** 0.428*** 0.379*** 0.420*** (0.015) (0.017) (0.016) (0.017) (0.015) Constant -0.178*** -0.111*** 0.092*** -0.003 -0.121*** (0.015) (0.015) (0.015) (0.014) (0.019) Sample size 23591 23572 23591 23591 23572 R2 0.182 0.148 0.16 0.128 0.193 34 Table II.3 IPW least squares regressions, interaction terms between various factors and country groups (outcome variable cutoff at 85%, 80%, and 90%) Family Schooling Poverty Status Urban Vs. Rural Disability Status Cut point 0.85 0.8 0.9 0.85 0.8 0.9 0.85 0.8 0.9 0.85 0.8 0.9 Highest educational level in the household (base category=No school) Primary 0.034*** 0.036*** 0.018* 0.028*** 0.029*** 0.013 0.026** 0.028** 0.01 (0.008) (0.009) (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) Junior secondary 0.100*** 0.095*** 0.091*** 0.069*** 0.062*** 0.062*** 0.067*** 0.060*** 0.060*** (0.010) (0.010) (0.009) (0.010) (0.010) (0.009) (0.010) (0.010) (0.009) Senior secondary or higher 0.116*** 0.118*** 0.103*** 0.080*** 0.081*** 0.069*** 0.081*** 0.082*** 0.071*** (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) No School 0.054*** 0.056*** 0.049*** (0.011) (0.011) (0.010) No School#Mid-reading country -0.163*** -0.169*** -0.125*** (0.015) (0.016) (0.014) No School#High-reading country -0.102*** -0.088*** -0.129*** (0.024) (0.024) (0.024) Wealth index (base category=Poorest) Second quintile 0.043*** 0.042*** 0.042*** 0.040*** 0.039*** 0.037*** 0.043*** 0.042*** 0.041*** (0.010) (0.010) (0.009) (0.010) (0.010) (0.009) (0.010) (0.010) (0.009) Middle 0.074*** 0.079*** 0.072*** 0.070*** 0.075*** 0.065*** 0.074*** 0.079*** 0.071*** (0.010) (0.010) (0.009) (0.010) (0.010) (0.009) (0.010) (0.010) (0.009) Fourth quintile 0.142*** 0.144*** 0.128*** 0.137*** 0.139*** 0.120*** 0.140*** 0.143*** 0.124*** (0.011) (0.011) (0.010) (0.011) (0.011) (0.010) (0.011) (0.011) (0.010) Richest 0.253*** 0.261*** 0.238*** 0.245*** 0.254*** 0.225*** 0.247*** 0.256*** 0.227*** (0.012) (0.012) (0.012) (0.012) (0.012) (0.012) (0.012) (0.012) (0.012) Poor -0.043*** -0.046*** -0.037*** (0.012) (0.012) (0.010) Poor#Mid-reading country -0.083*** -0.087*** -0.064*** (0.016) (0.017) (0.015) Poor#High-reading country -0.045* -0.037 -0.084*** (0.022) (0.022) (0.021) Location (base category: urban) -0.124*** -0.120*** -0.114*** -0.185*** -0.185*** -0.168*** -0.075*** -0.081*** -0.045*** -0.115*** -0.112*** -0.106*** (0.009) (0.009) (0.009) (0.008) (0.008) (0.008) (0.013) (0.013) (0.012) (0.009) (0.009) (0.009) Rural#Mid-reading country -0.029 -0.014 -0.047** (0.017) (0.017) (0.016) Rural#High-reading country -0.117*** -0.104*** -0.164*** (0.018) (0.018) (0.018) Disabled (base category: non-disabled) -0.156*** -0.164*** -0.140*** -0.169*** -0.178*** -0.153*** -0.158*** -0.167*** -0.143*** -0.131*** -0.142*** -0.111*** (0.015) (0.015) (0.014) (0.015) (0.015) (0.014) (0.015) (0.015) (0.014) (0.021) (0.021) (0.019) Disabled#Mid-reading country -0.03 -0.029 -0.026 (0.027) (0.028) (0.025) Disabled#High-reading country -0.053 -0.041 -0.092* (0.042) (0.041) (0.039) 35 Age (Base category=10) age11 0.063*** 0.063*** 0.053*** 0.064*** 0.063*** 0.054*** 0.063*** 0.062*** 0.053*** 0.062*** 0.061*** 0.052*** (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) age12 0.112*** 0.117*** 0.092*** 0.111*** 0.116*** 0.092*** 0.112*** 0.117*** 0.092*** 0.111*** 0.117*** 0.092*** (0.009) (0.009) (0.008) (0.009) (0.009) (0.009) (0.009) (0.009) (0.008) (0.009) (0.009) (0.008) age13 0.163*** 0.172*** 0.135*** 0.163*** 0.173*** 0.135*** 0.162*** 0.171*** 0.134*** 0.162*** 0.171*** 0.134*** (0.009) (0.009) (0.008) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) age14 0.207*** 0.218*** 0.174*** 0.210*** 0.221*** 0.177*** 0.207*** 0.218*** 0.174*** 0.206*** 0.217*** 0.173*** (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) Gender (Base category: Men) 0.038*** 0.038*** 0.034*** 0.040*** 0.040*** 0.035*** 0.038*** 0.038*** 0.034*** 0.038*** 0.038*** 0.034*** (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) Country 0.314*** 0.337*** 0.238*** 0.307*** 0.330*** 0.237*** 0.293*** 0.305*** 0.238*** 0.279*** 0.301*** 0.214*** (0.008) (0.009) (0.008) (0.009) (0.009) (0.008) (0.014) (0.014) (0.013) 0.008 0.008 0.007 0.478*** 0.483*** 0.430*** 0.457*** 0.460*** 0.418*** 0.522*** 0.520*** 0.502*** 0.456*** 0.461*** 0.410*** (0.009) (0.009) (0.010) (0.010) (0.010) (0.010) (0.013) (0.013) (0.014) 0.009 0.009 0.009 Constant 0.009 0.017 -0.004 0.113*** 0.126*** 0.094*** -0.040* -0.026 -0.058*** -0.019 -0.01 -0.025 (0.016) (0.016) (0.016) (0.014) (0.015) (0.014) (0.017) (0.017) (0.016) 0.016 0.016 0.016 Sample size 23572 23572 23572 23572 23572 23572 23572 23572 23572 23572 23572 23572 R2 0.208 0.215 0.175 0.192 0.198 0.163 0.207 0.214 0.178 0.206 0.212 0.175 36 Table II.4 IPW least squares regressions, interaction terms between disability status and social factors (outcome variable cutoff at 85%, 80%, and 90%) Cut point 0.85 0.8 0.9 Disabled (base category: non-disabled) -0.249*** -0.268*** -0.249*** (0.045) (0.047) (0.040) Location (base category: urban) -0.117*** -0.114*** -0.103*** (0.009) (0.009) (0.009) Disabled # Location Disabled # Rural 0.055 0.054 0.092** (0.035) (0.035) (0.032) Wealth index (base category=Poorest) Middle 0.081*** 0.081*** 0.076*** (0.008) (0.009) (0.008) Richest 0.232*** 0.238*** 0.218*** (0.012) (0.012) (0.012) Disabled # Wealth Index Disabled#Middle 0.000 0.02 -0.003 (0.034) (0.035) (0.031) Disabled#Richest 0.01 0.039 0.013 (0.057) (0.056) (0.053) Primary 0.034*** 0.032*** 0.024** (0.009) (0.009) (0.008) Junior secondary 0.109*** 0.105*** 0.092*** (0.010) (0.010) (0.010) Senior secondary or higher 0.096*** 0.095*** 0.084*** (0.011) (0.011) (0.011) Disabled # Highest Education level in the household Disabled#1 0.094** 0.086* 0.058 (0.036) (0.037) (0.031) Disabled#2 0.049 0.056 0.041 (0.039) (0.041) (0.034) Disabled#3 0.073 0.048 0.098* (0.050) (0.051) (0.046) Age (Base category=10) age11 0.064*** 0.063*** 0.054*** (0.009) (0.009) (0.009) age12 0.111*** 0.116*** 0.092*** (0.009) (0.009) (0.008) age13 0.162*** 0.172*** 0.136*** (0.009) (0.009) (0.008) age14 0.206*** 0.217*** 0.174*** (0.009) (0.009) (0.009) Gender (Base category: Boys) 0.036*** 0.036*** 0.032*** (0.006) (0.006) (0.006) Country Chad 0.065*** 0.070*** 0.068*** (0.017) (0.017) (0.015) DRCongo 0.043** 0.053*** 0.037** (0.014) (0.014) (0.013) Ghana 0.315*** 0.332*** 0.268*** (0.015) (0.015) (0.014) Lesotho 0.466*** 0.487*** 0.360*** (0.017) (0.017) (0.017) Madagascar 0.388*** 0.431*** 0.304*** (0.016) (0.016) (0.015) Malawi 0.379*** 0.413*** 0.289*** (0.014) (0.015) (0.014) The Gambia 0.221*** 0.231*** 0.187*** (0.018) (0.018) (0.016) Togo 0.270*** 0.280*** 0.208*** (0.017) (0.017) (0.015) Tunisia 0.698*** 0.723*** 0.636*** (0.014) (0.014) (0.015) Zimbabwe 0.427*** 0.424*** 0.414*** (0.015) (0.015) (0.015) _cons -0.097*** -0.093*** -0.100*** (0.019) (0.019) (0.018) Sample size 23572 23572 23572 R2 0.222 0.233 0.19