Quality and inequality in pre-primary and home environment inputs to early childhood development in Egypt

By the time children start primary school, large socioeconomic disparities are evident in their learning and development. Both pre-primary and home environments can play important roles in influencing school readiness and can contribute to disparities in early childhood development, but there is limited evidence on their relative roles in the Middle East and North Africa. This paper examines how pre-primary quality, stimulation at home, and early childhood development vary by socioeconomic status for pre-primary students in Egypt. The results demonstrate substantial socioeconomic inequality in stimulation at home, more so than in pre-primary quality and inputs, although there is variation in the degree of inequality across different dimensions of pre-primary quality. “ Double inequality ” is observed, where students with less stimulating home environments experience slightly lower quality pre-primary inputs. There are particularly large pre-primary inequities in structural quality (physical environment) and less inequity in process quality (pedagogy). These results suggest that targeted investments in pre-primary education in Egypt are necessary to reduce inequality in school readiness but are likely insufficient to close the socioeconomic status gap in children ’ s development. Investing in interventions to improve vulnerable children ’ s home learning environments, as well as investing in quality pre-primary, is critical to address disparities in children ’ s development.


Introduction
Education systems in low-and middle-income countries (LMICs) struggle with a learning crisis.While school enrollments have expanded, learning and even basic skills such as literacy lag for children in LMICs compared to high-income countries (HICs) (Le Nestour, Moscoviz, & Sandefur, 2022;World Bank, 2018).A key reason that children in LMICs fall behind is underinvestment in early childhood development (ECD), generally, and pre-primary education specifically (Nores & Barnett, 2010;World Bank, 2018).
Millions of children are at risk for low-quality care in the early childhood years (McCoy, Seiden, Cuartas, Pisani, & Waldman, 2022).Additionally, substantial disparities in stimulating home environments and access to pre-primary across and within LMICs have been documented (Krafft & El-Kogali, 2021;McCoy et al., 2018).While there is a large literature documenting inequality in ECD and home environments in both HICs and LMICs (Flood, McMurry, Sojourner, & Wiswall, 2022;Walker et al., 2011), and research exists on unequal pre-primary quality in HICs (Flood, McMurry, Sojourner, & Wiswall, 2022), less is known about inequities in access to quality pre-primary learning environments in LMICs.
Pre-primary education can particularly benefit disadvantaged children (Holla, Bendini, Dinarte, & Trako, 2021) by reducing gaps in learning and development that form before the start of formal schooling.The recent push to increase pre-primary enrollments in LMICs recognizes the importance of the quality of pre-primary programs (Bendini & Devercelli, 2022).However, there is little research on whether the disadvantaged children who do attend pre-primary in LMICs have equal access to quality pre-primary environments.Inequality in pre-primary quality could be more, less, or similar to inequalities in home environment quality in LMICs.The relative degree of inequality in home and pre-primary environments has important implications for the potential of pre-primary to reduce, maintain, or exacerbate school readiness gaps for disadvantaged children.
This study uses data from kindergartens (KGs) and KG students in the Arab Republic of Egypt to investigate quality and inequality in both pre-primary and home environmentsthe two central drivers of ECD for pre-primary students.It is particularly unusual to have data on both pre-primary quality and home environments in LMICs, to be able to examine inequality as well as potential complementarities or substitutions between these important inputs.Egypt is a valuable setting to be able to assess this inequality; the country has relatively low pre-primary enrollments compared to other countries at similar levels of development (El-Kogali & Krafft, 2015).Pre-primary is also the phase of education in Egypt with the largest socio-economic inequality (Krafft & El-Kogali, 2021).
Based on existing literature, we hypothesized that disparities in the quality of learning environments (at home and in pre-primary settings) and disparities in children's developmental outcomes would be evident in Egypt, as they are in other countries.We specifically examine, for Egyptian pre-primary students, how pre-primary quality, stimulation at home, and early childhood development outcomes vary by socio-economic status (SES).We further explore which aspects of pre-primary quality (such as the physical environment or pedagogy) are more equitable.Due to lack of research on young children in Egypt, we had no specific hypotheses on whether disparities would be larger in home or pre-primary learning environment, nor did we hypothesize on the relative contributions of each to young children's developmental outcomes.We undertake these analyses using factor analysis to generate measures of ECD, dimensions of pre-primary quality, and stimulation at home.We then use regression models for how these outcomes relate to SES.
The results demonstrate that pre-primary students face substantial socio-economic inequality in stimulation at home, more so than in pre-primary quality and inputs.We find, for example, that pre-primary students with mothers with no education experience an average of 0.84 standard deviations lower level of stimulation at home than children of university-educated mothers.Inequality in students' experiences of quality pre-primary education varies substantially by the dimension of quality in question; we document the largest inequities in structural quality (physical pre-primary environment) and less inequity in process quality (teacher practices, children's experience of quality materials and adherence to the curriculum).These results suggest that preprimary classrooms may provide relatively more equal opportunities for children to learn, and thus may have an important role to play in reducing the school readiness SES gap.At the same time, the importance of the home learning environment in shaping developmental outcomes, which varies substantially by SES, is profound.Very high-quality learning environments are likely required for the most disadvantaged children to begin to address the gaps in learning that are evident in pre-primary due to home learning environments.Yet, even the highest quality preprimary learning environment may be insufficient to address disparities in early development, indicating the need to address inequalities in home environments as well.

Evidence on inequality in early childhood development
Data from across the globe emphasize important gaps in development that start in early childhood and persist thereafter (Britto et al., 2017;McCoy, Seiden, Cuartas, Pisani, & Waldman, 2022).For instance, at age five, children in Peru, Ethiopia, India, and Vietnam all already had large socioeconomic gaps in their vocabulary (Lopez Boo, 2016).Some evidence suggests that disparities in early cognitive development emerge within the first year of life (Fernald, Kariger, Hidrobo, & Gertler, 2012).Across Cambodia, Mongolia, and Vanuatu, there were significant SES gradients in language, literacy, mathematics, and executive function of children aged 3-5 years, with preprimary attendance mediating some of the disparities (Sun, Zhang, Chen, Lau, & Rao, 2018).Both home and pre-primary environments can play an important role in ECD and early inequality.

Evidence on the importance of a stimulating home environment
In LMICs, both nutrition and responsive caregiving (a stimulating home environment) have been shown to be critically important to children's subsequent development in the first years of life (Britto et al., 2017;Gertler et al., 2014).A substantial body of research from LMICs documents the importance of parenting for children's development in early childhood, both concurrently and longitudinally (Knauer, Ozer, Dow, & Fernald, 2019;Lu et al., 2020).Programs promoting improvements in parenting can change behaviors in ways that improve children's development, although not necessarily all dimensions of development (Premand & Barry, 2020).Early childhood interventions with an educational or stimulation component had the largest cognitive effects in a review of the international literature (Nores & Barnett, 2010).However, children do not have equal access to stimulating home environments, contributing to inequality in ECD and continuing throughout childhood and adolescence.For example, in Brazil and South Africa, children with higher-quality home learning environments during the pre-primary years had higher IQs and greater psychosocial adjustment as young adults (Trude et al., 2021).These early inequities in stimulating home environments arise from a number of environmental influences, starting at preconception and extending throughout early childhood, including lack of access to adequate health care and nutrition; lack of social protection; and poverty (Britto et al., 2017).Early gaps between children with supportive environments and those with less supportive environments persist over time, leading to substantial inequities at the start of formal schooling that persist throughout childhood (Crosnoe, Leventhal, Wirth, Pierce, & Pianta, 2010) and may even fully account for SES-related group differences in early developmental outcomes (Rosen et al., 2020).

Evidence on the importance of quality pre-primary teaching and learning
Inequality in pre-primary education starts with whether children are able to attend pre-primary at all.Family wealth and maternal education have been shown to have a powerful influence on whether young children access early childhood education (Rao, Cohrssen, Sun, Su, & Perlman, 2021), compounding the impact of home environments.Higher-income and more educated parents are more likely to ensure their children attend high quality early childhood settings, which in turn leads to compounding disparities as children with more home stimulation are also more likely to have stimulating out-of-home learning environments (Alexandersen, Zachrisson, Wilhelmsen, Wang, & Brandlistuen, 2021;Meyers & Jordan, 2006;Rao, Cohrssen, Sun, Su, & Perlman, 2021).
Even when children are able to attend pre-primary education, existing research suggests that highquality 5 pre-primary settings may be relatively rare within LMICs.While there are few sources of descriptive data on pre-primary classrooms in LMICs, research suggests that classrooms may lack child-centered approaches and access to materials that promote children's learning, with high reliance on rote instruction of early academic skills and unsafe physical conditions (Bidwell & Watine, 2014;Raikes, Koziol, Davis, & Burton, 2020).While early childhood care and education can have a positive impact on school readiness and reduce inequality (Jung & Hasan, 2014;Krafft, 2015;Nores & Barnett, 2010;Nores, Bernal, & Barnett, 2019), in some studies, attending lowquality pre-primary led to no or negative effects on children's development compared to alternative care or schooling arrangements (Blimpo, Carneiro, Jervis, & Pugatch, 2019;Bouguen, Filmer, Overall, evidence from HICs demonstrates that attending pre-primary education, especially highquality early childhood care and education, both improves school readiness and particularly benefits disadvantaged groups (Heckman, 2006;Magnuson, Meyers, Ruhm, & Waldfogel, 2004;Temple & Reynolds, 2007).However, effect sizes linking quality environments with child development are small (e.g., Brunsek et al., 2017;Perlman et al., 2016), highlighting both measurement challenges in capturing the complexity of learning environments (Burchinal, 2018) and multiple influences on children's development.Existing work in many LMICs demonstrates associations between aspects of quality in pre-primary settings and ECD (Aboud, Hossain, & O'Gara, 2008;Andrew et al., 2022;Brinkman et al., 2017;McCoy & Wolf, 2018;Mwaura, Sylva, & Malmberg, 2008;Raikes, Koziol, Davis, & Burton, 2020;Rao, 2010;Rao, Richards, Sun, Weber, & Sincovich, 2019).But similar to HICs, research in LMICs has demonstrated small but significant associations between pre-primary quality and ECD, albeit with many of the same challenges noted in research on HICs (Chen & Wolf, 2021).Given the stronger link between home environments and ECD than pre-primary quality and ECD in the literature, the quality and inequality of these inputs may have differential effects on developmental trajectories.

Egypt's pre-primary system
At age four in Egypt, children are eligible for KG, which serves children aged 4-6.KG is not compulsory, and children can enter at either KG 1 or KG 2. The KG school-day averages five hours. 7The Ministry of Education and Technical Education (MoETE) oversees KGs and provides public KG classes in public primary schools.The majority of KG enrollment is in the public sector, with private provision at 26%.8 Private KGs are primarily attended by children from wealthy households (El-Kogali & Krafft, 2015).There is substantial inequality in pre-primary enrollment in Egypt in general, with children from wealthier, more educated households more likely to attend pre-primary (El-Kogali & Krafft, 2015;Krafft, 2015;Krafft & El-Kogali, 2021).For instance, only 16% of children from the poorest quintile of households attended pre-primary, compared to 65% of children from the richest quintile of households.Likewise, only 20% of children with mothers who had no education attended pre-primary compared to 65% of children with mothers with higher education (El-Kogali & Krafft, 2015).
Pre-primary enrollment in Egypt has historically been substantially below the world average but has recently been rising (Figure 1).Around 2000, the pre-primary gross enrollment rate hit 10%, and reached 28% as of 2010 but then plateaued. 9The "echo" of the youth bulge, the result of the youth bulge entering childbearing age compounded by a rise in fertility in the early 2010s, placed demographic pressure on Egypt's pre-primary and primary education system (Assaad, 2020;Krafft, 2020;Krafft & Assaad, 2014;Krafft, Assaad, & Keo, 2022).

Education challenges in Egypt
Generally, investment in early childhood development in the MENA region has been comparatively low (El-Kogali & Krafft, 2015).The region also tends to have the lowest scores in the world on international assessments during the primary and secondary grades (El-Kogali & Krafft, 2020).The catch-up between the earlier grades of primary school and later grades suggests that gaps in early learning may play an important role in learning and human capital challenges in the region (El-Kogali & Krafft, 2020).

Education 2.0 reforms
Starting in 2018, MoETE began a series of system-wide educational reforms, referred to as education 2.0 (Moustafa, Elghamrawy, King, & Hao, 2022).The new education 2.0 system was competency-based, multi-disciplinary, and aimed to foster a variety of 21 st century skills.The new system also included a new approach to assessment and examination.Goals of the reform included expanding access to pre-primary education and improving the quality of education.Reforms were implemented grade by grade, starting with the pre-primary level (Moustafa, Elghamrawy, King,  & Hao, 2022). 10The project was supported in part by a World Bank loan, including $80 million focused on pre-primary quality (World Bank, 2022b).
A new curriculum, textbooks, and teacher guides were developed, along with complementary online content (Moustafa, Elghamrawy, King, & Hao, 2022).The new materials for KGs were quite detailed, laying out a daily schedule,11 providing teaching strategies, and detailed, scripted instructions in the teacher's guide (Ministry of Education and Technical Education, 2019).12Children had corresponding workbooks.Activities often required corresponding materials, e.g.manipulatives for counting.Continuous professional development was planned, with initial training on the new education system and materials offered and ongoing training planned.Training was primarily offered in a cascade model (Moustafa, Elghamrawy, King, & Hao, 2022).The standard amount of professional training was three days per term.
3 Hypotheses, data, and methods

Hypotheses
We hypothesize that there are disparities in children's access to quality environments (in-home and out-of-home) as well as in child development.We have three specific hypotheses: 1) Egyptian pre-primary students have unequal ECD that depends on their SES, with higher SES children demonstrating more developed skills and competencies.
2) Egyptian pre-primary students have experienced unequal home environments that depend on their SES, with higher SES children experiencing more stimulating and supportive home learning environments.
3) Egyptian pre-primary students have unequal pre-primary environments that depend on their SES, with higher SES children experiencing higher quality pre-primary learning environments.
We test these hypotheses for outcomes based on factor analyses for ECD, pre-primary quality and its dimensions, and home stimulation.We use data from a sample of KGs designed to be nationally representative and assess the magnitude and statistical significance of relationships between outcomes and SES using descriptive approaches and regression models, as detailed below.

Data collection tools, training, and fieldwork
The Measuring Early Learning Quality Outcomes (MELQO) tools (UNESCO, 2017) were the foundation of data collection, locally adapted to the Egyptian context.The MELQO tools have two main components, the Measure of Early Development and Learning (MODEL) for measuring the development of children aged 3-6, and the Measure of Early Learning Environments (MELE).The MODEL collects data through a child direct assessment, parent report of child development (including home and family background), and teacher report of child development.The MELE collects data via classroom observation, a teacher interview, parent interview, and school director interview.The tools were designed specifically to measure child development and quality of early childhood education in low-and middle-income countries (Raikes et al., 2019).
The MELQO tools were developed initially and piloted in 2015 in non-representative samples, and then the pilot-tested tools were used in national studies starting in 2016.The tools were finalized and publicized in 2017 (Raikes et al., 2019).The tools were selected for a nationally representative study of kindergarten teaching and learning in Egypt.The primary goal of the Egyptian data collection was to identify quality issues and inform subsequent teacher training efforts for Education 2.0, as well as to validate a new quality assurance system.
The MELQO tools were translated into Arabic and adapted to the Egyptian context and curriculum in collaboration with the MOETE, kindergarten teachers, and kindergarten supervisors.An adaptation workshop occurred in May 2019 that included a careful review of items by a group of stakeholders along with addition or modification of items to align with national standards and cultural priorities (for example, items specifically focused on the implementation of Egypt's new standards for pre-primary classrooms).The tools were programmed into Android tablets using ODK-X software (Brunette et al., 2017).Pre-piloting of the instruments subsequently took place in Egypt in two governorates, ten schools, ten classrooms, with ten teachers and 30 children.
Training of the master trainers, a mix of MOETE officials, supervisors, and Egyptian academic experts, by the international experts took place in January 2020.Training of enumerators took place over 10 days starting in late February 2020, including piloting in schools.Enumerators were required to reach scores of at least 80% on activities and quizzes during training, to ensure adequate inter-rater reliability.Enumerators were graduates of faculties of kindergarten education or child psychology, or kindergarten teachers or supervisors.Data collection was initially scheduled to take place in mid-March 2020.On the date data collection was supposed to begin, schools were closed due to COVID-19.
In Fall 2021, public schools reopened on October 9.After schools opened, a repeat of training was held for enumerators.Data collection in schools took place from November 6, 2021, to December 8, 2021.Parents were interviewed over the phone through December 15, 2021.

Sample
The study sample was designed to be nationally representative of Egyptian KGs and their students.Egypt's Education Management Information System (EMIS) database from 2018-19 was the sample frame.The sample was stratified by type (public versus private), region, 13 and community poverty status. 14Within each public/private, region, and poverty status strata, a random sample totaling 46 districts was drawn. 15Five schools were randomly selected within each district. 16A total of 214 schools were sampled.17Data from 213 schools are included in this study. 18ata were collected for up to three KG1 and three KG2 classes per school (randomly selected if more than three).There were 638 classrooms with child and teacher data completed.A random sample of four children per classroom was selected.The sample of children whose data were successfully collected was 2,455 observations. 19The data collection firm tried up to three times to reach parents, based on phone numbers provided by the school.For the parent data, there was substantial non-response (primarily that parents did not pick up, but some refusal when reached) such that only 1,437 parents were reached and consented. 20We focus on the sub-sample with parental data in order to be able to investigate home environments and inequality.

Outcomes
We examine three main categories of outcomes: early childhood development (collected through direct assessments and teacher reports), pre-primary quality (collected through observations), and stimulation at home (collected through parent reports).We summarize a large number of variables (as detailed in the appendix) into factors using confirmatory factor analysis. 21The only selection criteria was that the first factor has an eigenvalue of at least one.We kept even items with low loadings in making the index, but since the loadings were small, they have a small role in determining the value of the factor.
For ECD, we examine: For stimulation at home, we examine, based on the parent reports, children's books at home, and in the past 7 days engaging in the following activities: reading at home, singing songs, playing, telling stories.

Covariates
We control for child sex and the child's age in months in our models.In terms of family background, we include a number of items we refer to in brief as socio-economic status (SES).An asset index based on a factor analysis of owning various durable goods and housing conditions is included in the SES domain.Data on mother (or female caregiver) and father (or male caregiver) education level, along with corresponding parental occupation was also included in the domain of SES.We describe the characteristics of our sample in terms of mother and father characteristics in the appendix.

Methods
We undertake confirmatory factor analysis to generate our key outcomes.We provide details on the factor analyses in the appendix, and illustrative examples in the body of the paper.
We present descriptive statistics on inequality in KG students' development, pre-primary quality, and home stimulation by SES.We use visualizations of mean outcomes by mother's and father's characteristics and local polynomials (using a triangle kernel) of outcomes relative to the continuous asset index.In additional descriptives we show how stimulation at home and different aspects of pre-primary quality are related (also using local polynomials), highlighting how the different inputs to ECD can potentially offset or compound inequality in ECD.
We estimate a series of ordinary least squares (OLS) models for these different outcomes including SES.Denote the outcome for child i as Yi.Denote the covariates as MEi,j for mother's education, FEi,j for father's education, MOi,j for mother's occupation, FOi,j for father's occupation, Ai for the asset index, Si for child sex, and Ci for child's age in months.We thus estimate: We cluster standard errors on the school level.Since different aspects of SES are likely to be multicollinear, and since we are testing a number of individual covariates, we also undertake tests for the joint significance of the categorial SES variables (mother's education; father's education; mother's occupation; father's occupation).
Weights are used in all our analyses.The weights incorporate the original sampling design, including on the school level, random sampling of classes, and random sampling of students.
Weights also account for non-response.Non-response accounts for the number of observations that should have been included (for example, the number of children or parents per class or per school).

Examples of outcomes and inputs
In Figure 2 we provide examples of ECD outcomes, home stimulation, and pre-primary inputs, in order to provide some sense of the ECD, home, and pre-primary context in Egypt.While only 46% of children reported being happy while reading in the direct assessment, 55% are always or often considerate per the teacher report, 62% of the time children correctly recognize letters in the direct assessment, 71% of the time they had accurate forward digit span (direct assessment), and 87% of the time they could count to ten (direct assessment).

Figure 2. Examples of early childhood development (ECD) outcomes and home and preprimary inputs
Source: Authors' calculations In terms of home stimulation, parents were asked how many days in the last 7 (from 0-7) someone in the household engaged in various activities with the child.Reading was rarest (1.3 days on average), followed by telling stories (1.7 days), singing (1.8 days) and then most frequently playing (5.8 days).Data from pre-primary observations revealed that 46% of children attended a preprimary with at least one physical hazard and 67% attended a pre-primary where the teacher agreed or strongly agreed they were overwhelmed by their work.Although only 56% had a portfolio to track their development, 81% of children were in classes where children received individual instruction during the observation, and in 83% of cases the preparation record matched the schedule in the teacher's guide.The results exhibit meaningful variation; while some children are achieving key ECD benchmarks and experiencing high-quality inputs, others are not.

Inequality in early childhood development outcomes
In this section, we substantiate inequality in ECD by SES (testing H1).We examine the language, math, executive function, socio-emotional, and overall school readiness ECD outcomes and how they vary by SES. Figure 3 presents the patterns of the various ECD development outcomes by the asset index, based on a local polynomial (triangle kernel).Figure 4 shows ECD outcomes by mother's and father's education and father's occupation (few mothers work).Table 1 shows multivariate models of how ECD outcomes depend on SES, controlling for child sex and age.
Within specific domains of child development and across all domains there is a clear socioeconomic gradient in ECD (consistent with H1).Descriptively, there is a very similar pattern for all outcomes.In the multivariate models, the magnitude of the relationship is relatively similar; a one SD increase in the asset index predicts between a 0.111 and 0.171 SD increase in the ECD outcome, depending on the outcome (consistent with H1).All are statistically significant at the 5% level except for executive functioning (0.111).

Figure 3. Child outcome factors (in standard deviations [SD]) by asset index (in SD)
Source: Authors' calculations Notes: Local polynomial with triangle kernel, bandwidth two.Visualizing from 1 st -99 th percentile.
There are particularly large differences in child outcomes by mother's education (consistent with H1).Descriptively (Figure 4), children of mothers reporting no formal education have scores on the school readiness factor of -0.61 (factors are normalized, so factors are measured in standard deviations), compared to -0.16 for mothers with vocational secondary (the most common degree in Egypt (Krafft, Assaad, & Keo, 2022), although university is the most common in the sample, since enrollment in pre-primary is itself unequal).Only at the university level is readiness above average (0.25).
In the multivariate models, mother's education is jointly significant in predicting math, executive functioning, and overall school readiness skills (consistent with H1).Compared to a mother with no education, a mother with university education predicts an 0.555 SD higher overall readiness factor.There are similar but somewhat smaller descriptive disparities by fathers' education, father's occupation, and mother's occupation (which are all highly correlated with mother's education and other aspects of SES).None of the categories is jointly significant in any of the models.Overall, there are clear socio-economic disparities (consistent with H1), most closely related to mother's education, but also tied to household wealth and income (proxied by the asset index).

Inequality in inputs
We now turn to examining inequality in home and pre-primary inputs by SES.In Figure 5, we explore the patterns of pre-primary quality and home inputs by the families' asset index, based on a local polynomial (triangle kernel).Figure 6 shows the variation in stimulation by mother's and father's education and father's occupation.Table 2 shows OLS models for SES and the various home and pre-primary inputs (testing H2 and H3).There is substantial variation in the relationship between inputs and assets.Strong relationships were observed between home stimulation and preprimary environments, and the family asset index.A one SD increase in the asset index predicts a statistically significant 0.197 SD increase in home stimulation (consistent with H2).There are similar (and likewise significant) relationships of around 0.19 SD increases in the pre-primary environment or teacher attitudes for each SD increase in assets (consistent with H3).Other results for teaching practices (coefficient of 0.117), materials (-0.006) and adherence to the curriculum (-0.069) were not significantly associated with family assets.

Figure 5. Input factors (in standard deviations [SD]) by asset index (in SD)
Source: Authors' calculations Notes: Local polynomial with triangle kernel, bandwidth two.Visualizing from 1 st -99 th percentile.
Although there are descriptive differences in a number of inputs by parent's characteristics (Figure 6), only a few are statistically significant.For instance, children of mothers with no education experience an average of a -0.57stimulation factor, compared to 0.27 for those with universityeducated mothers.Mother's education is jointly significant for home stimulation (consistent with H2) and teaching practices (consistent with H3, but only for this outcome) (Table 2).There are not significant differences for any of the inputs by father's education, using the joint tests.Mother's occupation is significant for teaching practices and curriculum adherence, but with children of mothers engaged in sales and service jobs having better outcomes than children whose mothers are in managerial/professional jobs.However, few mothers work at all.Father's occupation is only statistically significant for home stimulation, with all other statuses having significantly lower home stimulation (by -0.185 to -0.322 SDs) compared to managerial/professional fathers.
Although we have only one measure of home environment quality (stimulation at home), it is notable that we see stronger inequities in home environments than in pre-primary environments.
While different types of pre-primary inputs vary substantially in terms of their inequality, they are less unequal than home stimulation, particularly for materials and adherence to the curriculum, and to some extent teaching practices.In Figure 7 we specifically explore the relationship between home inputs (home stimulation) and pre-primary inputs, based on a local polynomial (triangle kernel).The correlations between home stimulation and pre-primary inputs are modest, with home stimulation not strongly correlated with quality of pre-primary environments.The strongest correlation (0.17) is with the environment, followed by teacher attitudes (0.13), materials (0.10) and teaching practices (0.08).Adherence to the curriculum is not correlated with stimulation at home (-0.02).Generally, students with more stimulating home environments are experiencing slightly higher quality pre-primary inputs.5 Discussion and conclusions

Summary
Using a representative sample, this study provides documentation of early disparities in children's learning outcomes and the quality of home and pre-primary learning environments.Our analyses demonstrate that early disparities documented in many countries are also evident in Egypt.We document disparities in children's learning outcomes in pre-primary.There are differences in children's language, math, executive function, socio-emotional, and overall school readiness outcomes by SES, particularly mother's education (consistent with H1).The role of mother's education may reflect substantial gender inequality in care work in Egypt (Economic Research Forum & UN Women, 2020;El-Feki, Heilman, & Barker, 2017), as mothers are typically the primary caregivers for children with much less direct involvement from fathers.
We also observe disparities in home learning environments (consistent with H2) and in preprimary quality (consistent with H3).There are socio-economic differences in children's home stimulation and components of their pre-primary education experience (including environment and teacher attitudes).Children who experience lower-quality home learning environments also experience lower-quality pre-primary education in some regards, but not all.Inequities are largest for structural quality (the pre-primary physical environment), whereas there is less inequity in process quality (teacher practices, children's experience of quality materials, and adherence to the curriculum).

Limitations
Our results indicate important disparities in ECD, home stimulation, and some aspects of preprimary quality and inputs that are critical to address.However, there are a number of limitations to our results that must be kept in mind and point to important areas for future work and research.First, we were only able to estimate correlations between SES, outcomes, and inputs.The causal effects of inputs, particularly pre-primary inputs, in LMICs are under-researched and an important area for future work.Second, we were comparing one measure of home stimulation to multiple dimensions of pre-primary quality.There may be other aspects of the home environment that we were not able to observe that are more or less unequal.Measuring quality of home stimulation or pre-primary learning environments is quite challenging, as is measuring the learning and development of young children (Burchinal, 2018).Ongoing efforts to improve measurement of ECD and early environments may reveal additional variation in inequality.Additionally, we do not know if one type of input (home or pre-primary, or a particular aspect of pre-primary quality) is more important than another in determining ECD.
Our analyses are based on a sample of pre-primary students.Not all children in Egypt attend preprimary; indeed, there is substantial socio-economic inequality in access to pre-primary (El-Kogali & Krafft, 2015).In the general population of pre-primary aged children (including those not attending pre-primary), there may be different patterns of inequality in home environments.The children not enrolled in pre-primary might particularly benefit from pre-primary or might particularly suffer from low-quality or inequitable pre-primary if they attended pre-primary; our research is not able to assess these dynamics, and they remain an important area for future research.
The sample we used from Egypt was designed to be nationally representative of pre-primary students, however, there was substantial non-response in the parental sample, which we use to measure SES.As Table 3, in the appendix, shows, there are some differences between our parental sample and a nationally representative sample of parents of KG students.The respondents in our sample were of slightly higher SES.This bias in the sample will not necessarily bias our research questions on SES unless there is a differential relationship among the respondents.
Our data collection efforts were also in late 2021, during the ongoing COVID-19 pandemic.While children were again attending pre-primary in person, the pandemic may have affected outcomes in complex ways that we are unable to unpack.These results do not necessarily generalize to other contexts, although future research should investigate the relative role of pre-primary and home environments in other LMICs.

Policy implications
Our findings point to two avenues for improving ECD and equity in ECD that can be pursued in parallel: First, investments in upgrading the pre-primary inputs that are relatively equal can help close ECD gaps for children who do attend pre-primary.For instance, since adherence to the curriculum is relatively equitable, improvements in curriculum quality may in turn lead to equitable improvements in ECD among pre-primary students.Equitable improvements will likely not, however, be sufficient to address the inequities in ECD that pre-date pre-primary and inequality in other pre-primary inputs.
Thus, second, targeted efforts should address the socio-economic inequality in both home and preprimary environments.Efforts must target children from less advantaged socio-economic backgrounds to ensure all children have equitable home environment, pre-primary, and ultimately ECD experiences.Although structural aspects of pre-primary quality may be easier for policy makers to standardize, they were more unequal than process components such as pedagogy.Addressing these structural inequities could help pre-primary better reduce gaps in school readiness for disadvantaged children.All these inputs should only be targets of policy inasmuch as they yield improvements in ECD.Although the literature suggests pre-primary quality and particularly the home environment matter for ECD, establishing which specific inputs have the highest causal impact on ECD within the Egyptian context would be valuable for informing policy.
Given the strong self-and cross-productivity of ECD skills (Helmers & Patnam, 2011), multidimensional inequality is likely to compound over time.Approaches to addressing learning poverty should likely focus on compensatory models that aim to provide extremely high-quality pre-primary education to children most at risk for poor ECD (which is the opposite of what we typically see in Egypt).Redressing inequality in early learning can not only improve outcomes and close gaps for disadvantaged students, it can also benefit their peers, improving learning for all (Berlinski, Busso, & Giannola, 2022).
However, the effects of pre-primary and pre-primary quality on school readiness and potentially compensating for inequitable home environments can be complex.For instance, an experiment in Mauritius showed that high quality pre-primary benefited children with low educated fathers, but led to worse outcomes for children with poorly educated mothers (Morabito, De, & Figueroa, 2018).Efforts to improve pre-primary quality and equity must carefully assess their actual impacts to determine the mix of interventions that will be most effective in closing gaps in early learning.
An important question that our research sheds light onbut cannot fully answeris whether preprimary or high-quality pre-primary can close school readiness gaps for disadvantaged children.
Children starting pre-primary already have unequal ECD due to unequal early home environments.
If pre-primary is substantially higher quality than home environments, even if it is somewhat unequal in quality, it could still close gaps.Moreover, if pre-primary quality is similar to home environment quality on average, and less unequal (this latter condition we have confirmed in Egypt), it could also help close gaps.
While we cannot directly estimate, in our work, the impact of pre-primary and quality pre-primary on ECD or the impact of improving home environments (e.g., early stimulation interventions) in LMICs, we can draw on the literature to assess the potential of pre-primary to close school readiness gaps.Effect sizes of pre-primary quality on learning in HICs tend to be around 0.1 if not smaller (e.g.Brunsek et al., 2017;Perlman et al., 2016).However, one recent meta-analysis found effect sizes of 0.25 on children's skills for interventions designed to improve pre-primary quality in HICs and 0.16 for pre-primary quality in LMICs (Holla, Bendini, Dinarte, & Trako, 2021).Quality improvements also had larger impacts than efforts to improve access (Holla, Bendini, Dinarte, & Trako, 2021).Interventions that improve home learning environments tend to have effect sizes in the 0.2-0.3range if not larger (Dong, Dong, Wu, & Tang, 2020;Knauer, Ozer, Dow, & Fernald, 2019;Zuilkowski, McCoy, Jonason, & Dowd, 2019).
As a point of reference, in Egypt, having a mother with no education versus a university education was associated with a raw readiness gap of 0.86 standard deviations.Trying to close the readiness gap with targeted pre-primary quality interventions alone would require a 5.4 standard deviation increase in pre-primary quality (using an effect size of 0.16 (Holla, Bendini, Dinarte, & Trako, 2021)).Improvements via home learning environments would require 2.9-4.3 standard deviation increases in home environments.These back-of-the-envelope calculations suggest targeted efforts towards both home environment and pre-primary quality are needed to help close school readiness gaps.

Areas for future research
Our findings point to several important areas for future research and data collection to inform policy.Nationally representative data at the pre-primary stage are rare in LMICs (Raikes, Sayre, & Lima, 2021), and data are important pre-requisite to evidence-based efforts to address inequality.Longitudinal data on young children and the trajectory of their development in LMICs are also much needed to understand critical points for intervention. 22urther research on promoting pre-primary quality and the impact of quality interventions on ECD is needed.Most of the evidence on what works to promote teaching quality and learning in LMICs comes from the primary level.For instance, only 8% of studies on education in Africa focused on pre-primary (Evans & Mendez Acosta, 2021).Research exploiting the random assignment of kindergarten students in Ecuador demonstrated that teachers' classroom practices are associated with higher learning (Araujo, Carneiro, Cruz-Aguayo, & Schady, 2016).At the pre-primary level, play-based learning can be particularly important and effective (Attanasio et al., 2019;Wolf, Aber, Behrman, & Tsinigo, 2019).However, play-based approaches can also face backlash from parents or teachers (Wolf, Aber, Behrman, & Tsinigo, 2019).
In addition, further research with rigorous causal identification strategies is needed to assess whether, when, and how pre-primary may help close gaps in ECD, as well as which specific input improvements would be most effective for improving equity and learning.Efforts to examine the impact of quality pre-primary on child development should therefore include estimates of the quality of children's home learning environments, given the large impact of home environments on children's learning and potential role of pre-primary and pre-primary quality in closing gaps.

Figure 4 .
Figure 4. Mean child outcome factors (in standard deviations [SD]) by parental education, father's occupation

Figure 6 .
Figure 6.Mean input factors (means, in standard deviations [SD]) by parental education and father's occupation

Figure 7 .
Figure 7. Pre-primary input factors (in standard deviations [SD]) by home stimulation factor (in SD)

Table 3 . Mother's and father's characteristics from KG sample and KG students in the Egypt Labor Market Panel Survey (ELMPS) 2018 sample
Source: Authors' calculations based on KG sample and Egypt Labor Market Panel Survey 2018