Received: 19October 2022 Revised: 28 June 2023 Accepted: 29 June 2023
DOI: 10.1111/desc.13434
R E S E A RCH ART I C L E
Quantifying quality: The impact of measures of school quality
on children’s academic achievement across diverse societies
Bruce S. Rawlings1 Helen Elizabeth Davis2,3 Adote Anum4 Oskar Burger5
Lydia Chen6 Juliet Carolina CastroMorales7 Natalia Dutra8
Ardain Dzabatou9 Vivian Dzokoto10 Alejandro Erut11 Frankie T. K. Fong12
Sabrina Ghelardi6 Micah Goldwater13 Gordon Ingram14 EmilyMesser15
Jessica Kingsford13 Sheina Lew-Levy1 KimberleyMendez6 MorganNewhouse6
Mark Nielsen16,17 Gairan Pamei18 Sarah Pope-Caldwell19 Karlos Ramos20
Luis Emilio Echeverria Rojas21 Renan A. C. dos Santos22 Lara G. S. Silveira22
JuliaWatzek23 CiaraWirth24 Cristine H. Legare11
1Department of Psychology &DurhamCultural Evolution Research Centre, DurhamUniversity, Durham, UK
2School of Human Evolution and Social Change & The Institute of HumanOrigins, Arizona State University, Tempe, Arizona, USA
Email: helenelizabethdavis@gmail.com
3Department of Human Evolutionary Biology, Harvard University, Cambridge, Massachusetts, USA
4Department of Psychology, University of Ghana, Accra, Ghana
5OMNI Institute, Denver, Colorado, USA
6Department of Psychology, The University of Texas at Austin, Austin, Texas, USA
7Keralty Lazos Humanos, Bogotá, Colombia
8Laboratório de Evolução do Comportamento Humano, Universidade Federal, Rio de Janeiro, Brazil
9Marien Ngouabi University Brazzaville, Brazzaville, Republic of the Congo
10Department of Psychology, Virginia Commonwealth University, Richmond, Virginia, USA
11Department of Psychology, Center for Applied Cognitive Science, The University of Texas at Austin, Austin, Texas, USA
12Max Planck Institute for Evolutionary Anthropology &, Department of Comparative Cultural Psychology, School of Psychology, University of Queensland, Brisbane,
Queensland, Australia
13School of Psychology, University of Sydney, Sydney, NSW, Australia
14Department of Psychology, Universidad de los Andes, Bogotá, Colombia
15Department of Psychology, Heriot-Watt University, Edinburgh, UK
16School of Psychology, University of Queensland, Queensland, Australia
17Faculty of Humanities, University of Johannesburg, Johannesburg, South Africa
18Department of Psychology, Chinese University of Hong Kong,Ma Liu Shui, Hong Kong
19Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
20The University of Texas at Austin, Austin, Texas, USA
21Independent Researcher, Unaffiliated
22Universidade Federal do Rio Grande, Rio Grande, Brazil
23Departments of Psychology & Philosophy, Neuroscience Institute, Georgia State University, Atlanta, Georgia, USA
24Tiputini Biodiversity Station, Colegio de Ciencias Biológicas y Ambientales, Universidad San Francisco deQuito USFQ, Quito, Ecuador
Bruce Rawlings andHelen Elizabeth Davis are the co-first authors and they contributed equally to this study.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2023 The Authors. Developmental Science published by JohnWiley & Sons Ltd.
Developmental Science. 2023;e13434. wileyonlinelibrary.com/journal/desc 1 of 17
https://doi.org/10.1111/desc.13434
2 of 17 RAWLINGS ET AL.
Correspondence
Bruce Rawlings, Department of Psychology & Abstract
DurhamCultural Evolution Research Centre,
DurhamUniversity, Durham, UK. Recent decades have seen a rapid acceleration in global participation in formal edu-
Email: bruce.rawlings@durham.ac.uk cation, due to worldwide initiatives aimed to provide school access to all children.
Research in high income countries has shown that school quality indicators have a
Funding information
National Science Foundation, Grant/Award significant, positive impact on numeracy and literacy—skills required to participate in
Number: 1730678; Templeton Religion Trust, the increasingly globalized economy. Schools vary enormously in kind, resources, and
Grant/Award Number: TRT0206
teacher training around the world, however, and the validity of using diverse school
quality measures in populations with diverse educational profiles remains unclear.
First, we assessedwhether children’s numeracy and literacy performance across popu-
lations improves with age, as evidence of general school-related learning effects. Next,
we examined whether several school quality measures related to classroom experi-
ence and composition, and to educational resources,were correlatedwith one another.
Finally, we examined whether they were associated with children’s (4–12-year-olds,
N = 889) numeracy and literacy performance in 10 culturally and geographically
diverse populations which vary in historical engagement with formal schooling. Across
populations, age was a strong positive predictor of academic achievement. Measures
related to classroom experience and composition were correlated with one another,
as were measures of access to educational resources and classroom experience and
composition. The number of teachers per class and access to writing materials were
key predictors of numeracy and literacy, while the number of students per classroom,
often linked to academic achievement, was not. We discuss these results in the con-
text of maximising children’s learning environments and highlight study limitations to
motivate future research.
KEYWORDS
cross-cultural comparison, formal education, global education, literacy, numeracy, school quality
RESEARCHHIGHLIGHTS
∙ We examined the extent to which four measures of school quality were associated
with one another, and whether they predicted children’s academic achievement in
10 culturally and geographically diverse societies.
∙ Across populations, measures related to classroom experience and composition
were correlated with one another as were measures of access to educational
resources to classroom experience and composition.
∙ Age, the number of teachers per class, and access to writing materials were key
predictors of academic achievement across populations.
∙ Our data have implications for designing efficacious educational initiatives to
improve school quality globally.
1 INTRODUCTION (Imchen & Ndem, 2020). The Sustainable Development Goals (Chasek
et al., 2016) represent the global standard for learning and formal edu-
Global participation in formal schooling has rapidly accelerated in cation and provide benchmarks for development initiativesworldwide.
recent decades. Since the early 1950s, the percentage of children A core goal of this initiative is to ‘ensure inclusive and quality edu-
worldwide attending primary school has risen from around 50%–92% cation for all and promote lifelong learning’. This emphasizes formal
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAWLINGS ET AL. 3 of 17
schooling, with the overall objective that ‘all girls and boys’ will com- turnaround) have provided promising recent data on school access
plete ‘free, equitable and quality primary and secondary education’ and quality, and children’s academic skill development. In a recent
by 2030. Despite this global increase in access to formal schooling, meta-analysis, school turnaround was associated with greater atten-
there remains significant inequity in school quality and student learn- dance andgraduation rates, aswell as improved scores on standardised
ing, however. Over half a billion children and adolescents worldwide tests (Redding & Nguyen, 2020). Studies outside these regions remain
are estimated to not reach minimum proficiency in numeracy and lit- scarce, however, and the limited research that exists generally focuses
eracy, and around 70%of children in low- andmiddle-income countries on years of schooling or highest grade level achieved, rather than edu-
cannot readorwrite at 10 years of age (WorldBank, 2022). This is com- cational quality or specific learned skills (Hanushek & Woessmann,
pared to 9%who do not meet this benchmark in high income countries 2012). Additional factors that have impeded research beyond high
(Imchen & Ndem, 2020). It is thus critical to understand how factors income populations include limited research on the sociocultural con-
such as school quality and access to education impacts skills such as texts of diverse populations. A particular challenge for global efforts
numeracy and literacy in populations with diverse socioeconomic and to expand educational access is that for many remote communities,
educational profiles. suchashunter-gatherer, pastoralist, and subsistence agricultural popu-
lations, there remain several obstacles to increasing school attendance
for children and caregivers who seek it. These include economic, polit-
1.1 The globalisation of formal education ical, social and cultural barriers, the stigmatisation children face by
peers at schools, and disparities between the structured and hier-
Children’s education has long been the focus of policies aimed at pro- archical approach of schools versus those of non-hierarchical and
moting future economic productivity and civic engagement (Smithers transitional communities (Ninkova et al., 2022; Siele et al., 2012). The
et al., 2018) and reducing poverty (Lim et al., 2018). Expanding educa- vast disparity in school quality between populations who have histor-
tional attainment in low-to-middle-income countries and low resource ically recent engagement with and access to formal education further
communities in high-income countries is an international priority, and emphasises the need to document how it impacts student learning of
within the last 50 years there has been a dramatic increase in access critical skills between populations.
to education. The globalization of formal education—defined as a Establishing reliable measures of school quality and their associa-
compulsory, structured education systemwhich typically follows a pro- tionwith literacy and numeracy is particularly important for three core
gramme or curriculum (UNESCO, 2011)—provides access to schools reasons. First, as noted, access to formal education is rapidly spread-
and educational resources to communities who historically have not ing across the globe. UNICEF aims to provide all children (∼3.5 billion
had exposure to this type of structured pedagogical environment. individuals) with access to digitised educational resources by 2030.
This expansion, including to remote populations, has stimulated cross- Second, educational philosophies vary dramatically across populations
cultural research to understand its impact on development (Gurven (Tobin et al., 2009). For instance, east Asian countries show a com-
et al., 2017; Legare et al., 2018). Of particular focus is literacy and paratively greater focus on rote learning compared to other nations
numeracy, which are increasingly considered critical skills in a global (Tan, 2011). Western educational philosophy tends to be individual-
labour market (Joynes et al., 2019). ist, and some non-western regions (e.g., sub-Saharan Africa) take a
Formal education has a particularly strong, positive impact on the more collectivist approach to education (Enslin & Horsthemke, 2016;
development of numeracy and literacy across development, as core Nsamenang, 2005). The extent to which school attendance is compul-
skills of focus in early education (Ball et al., 2014; Erbeli et al., 2021). sory and the frequency (e.g., number of hours per day and days per
In a growing number of communities worldwide, basic mathematics, week) in which children attend differs between nations—and in some
reading, and writing skills are critical to access the global economy cases, between regions and cities within nations. Whether variation in
and thus social mobility (Mok & Neubauer, 2016), as well as prosper- these types of factors shape learning of numeracy and literacy skills
ity and community integration (Ball et al., 2014; Tout, 2020). Research remains undocumented and thus poorly understood. Third, quality of
onWestern children shows that with age (and thus increased exposure schooling and educational resources also varies markedly. Variables
to formal education) children exhibit a strong increase in literacy and related to classroom experience and composition, such as the number
numeracy skills (Aunio & Niemivirta, 2010; Weinberger, 1996). How- of pupils and teachers per classroom, within-class grade ranges as well
ever, the relationship between age and academic achievement is less as teacher experience and attendance, all differ dramatically across
clear in some countries in the Global South, perhaps because many populations which have different access to formal education. Like-
children develop learning deficits early in schooling and fail to recover wise, factors associated with the availability of educational resources
(Spaull & Kotze, 2015). In addition, limited opportunities for develop- such as books, stationery or computers also vary substantially between
ing numeracy and literacy skills—an issue in many communities in the populations. This is particularly important because access to writing
Global South due to political, social, cultural, and economic conditions materials, such as pens, pencils, and notebooks, and lower student to
(Sepúlvedaet al., 2022)—is detrimental to children’s academic progress teacher ratios, improve children’s scholastic success because children
and future economic and vocational opportunities (Ball et al., 2014). can better engage in numeracy and literacy-based activities (Hungi &
Intensive short-term interventions aimed at dramatically improving Thuku, 2010) and have more direct teacher engagement (Francis &
the way a school operates in the US and Europe (often termed school Barnett, 2019).
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 of 17 RAWLINGS ET AL.
Another critical concern is that there are multiple ways to docu- TABLE 1 Sample composition by study population organised by
ment exposure to schooling and its impact on learning. This includes population size.
measures of attendance, performance on numeracy and literacy tasks, Gender (%
collating qualitative data (regarding, for example, attendance or atti- Age (years) boys)
tudes to education) from caregivers, children, and schoolteachers, as Population (units) Mean SD n (%)
well as school quality assessments (Humphreys et al., 2015). Since Manipur, India 8.88 2.50 93 54
understanding which factors significantly impact academic achieve-
Austin (TX), United States 7.45 1.99 62 47
ment across childhood requires reliable measures of school exposure
Natal, Brazil 8.98 2.06 118 50
and quality, it is crucial to assess the validity of difference measures,
San Cristobal, Colombia 9.64 1.68 100 43
and indeed whether measures related to classroom experience and
composition, as well as measures related to access to educational Keningau,Malaysia 8.66 2.01 131 55
resources are correlated with one another. Economic research into BlueMountains, Australia 9.62 1.87 66 52
the effects of education demonstrates that relying solely on years of Tanna, Vanuatu 8.27 2.27 101 53
schooling or highest grade level achieved in cross-cultural comparative Saltpond, Ghana 8.16 2.56 82 52
research does not provide an adequate understanding of the educa- Motaba River, Republic of 8.29 2.59 53 56
tional experience for most children worldwide (Lim et al., 2018), thus Congo
impeding the ability to assess variation in school quality in diverse Waoroni, Ecuador 8.57 2.35 83 55
educational contexts. Given the range of potential measures used in
research, studying school quality can thus be challenging. There are,
however, some recommendations that exist and that have been imple- on measures of literacy and numeracy by boys and girls across our
mented in research (Davis et al., 2020; Mayer et al., 2000; Singer & populations.
Braun, 2018). For example, as noted, access to classroom materials
(Davis et al., 2021), as well as classroom size for students and teach-
ers (Ehrenberg et al., 2001; Hanushek &Woessmann, 2012) have been 2 METHODS
linked to children’s performance in school (Francis & Barnett, 2019;
Hungi & Thuku, 2010). Participants included 889 children aged 4–12 years (M = 8.69,
The aim of this study was to assess the relations between school SD = 2.23, N = 456 boys) from 10 populations with diverse histori-
quality measures related to classroom experience and composition, cal engagement with and current access to formal education, including
and access to educational resources, and their impact on academic USA, Colombia, Brazil, Ecuador, Republic of the Congo, Ghana, India,
achievement in 4−12-year-old children across 10 geographically and Malaysia, Vanuatu, andAustralia. Thesedatawere collected as part of a
culturally diverse populations with diverse educational exposure and larger project investigating the influence of culture on children’s social
experiences. The populations that we studied here have historically learning and cognitive development (Burger et al., 2022; Dutra et al.,
variable access to formal schooling, with some having generations 2022).
of access to high quality schools, and others recent and limited Children were recruited in local communities or schools by
access to schools of variable quality. First, we assessed whether researchers in each community. Where local literacy rates allowed,
children’s numeracy and literacy performance across populations caregivers provided written informed consent prior to testing, and
improves with age, as evidence of general school-related learning where not, verbal consent was obtained. All participants provided
effects.We then examinedwhethermeasures of school quality related verbal assent directly before testing. All sites and methods were
to classroom composition and experience and amount of access to included in the ethical approval obtained from the University of Texas
educational resources were correlated. Finally, we examined whether at Austin Institutional Review Board (approval number 2017050101).
these measures of school quality predict numeracy and literacy The number of participants from each country ranges from 62 to
performance. 131 individuals, or 7% of the sample to 15% of the sample (Table 1).
We predicted that measures of school quality related to classroom Table 2 provides broader demographic and educational information
experience and composition (e.g., number of grades per school, the about each study population.
number of teachers and students per classroom) would be correlated
with one another. We also predicted that age, classroom composi-
tion and experience, and amount of access to educational materials 3 PROCEDURE
would be associated with higher academic performance among chil-
dren. Given that a large body of literature has reported considerable 3.1 Participant surveys
global variation by gender in access to schooling (Pekkarinen, 2012),
educational attainment (Legewie & DiPrete, 2012), and numeracy Childrenand/or their caregivers completeda survey capturingwhether
and literacy (Else-Quest et al., 2010; Voyer & Voyer, 2014), we con- and how long children attended school. All surveys were translated by
trolled for gender in all analyses, and we report overall performance local research assistants to local languages.
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAWLINGS ET AL. 5 of 17
TABLE 2 Demographic and educational information for the study sites. Literacy rates refer to national values of those aged 15 and over.
Population
size (Approxi-
Population mate) Population description Educational information by population
Manipur 2,850,000 Participants were children based in four districts of 98% of children enrol in school, and 62% secondary
(India) Manipur, India. Manipur is a state in the north-eastern school. Education is compulsory between 6–14
frontier of India borderingMyanmar. India is a large and years of age andmost children attend private
diverse country, however, the sampled districts of schools. Educationmaterials such as textbooks
Manipur– Imphal East, ImphalWest, Bishnupur, are almost exclusively in English, but children also
Churachandpur are representative of north-east India in learnManipuri andHindi. The national literacy
terms of the cultural commonalities, socio-economic rate is 77%. The primary language of instruction
levels, and educational contexts. The study sample differs across schools but is either English or
comprised ofMeeteis, Nagas, and Kukis which are the Meetei language. However, all school materials
three broad ethnic communities ofManipur. The latter are in English except for the language class of
two are predominantly Christians and are recognised by Manipuri (MeeteiMayek) or Hindi
the Indian government as indigenous tribes. Culturally
and linguistically, Manipur shares similarities with East
and Southeast Asian countries. Languages spoken by the
study sample aremostly tonal Tibeto-Burman languages,
but the primary ofManipur language isMeetei-lon. Our
sample consists of children from both private and public
schools in rural and urban areas and thus broadly
representative of other communities in the region
Austin (USA) 2,000,000 Austin is a largemetropolitan city and is the capital city of 95% of the population has completed primary
Texas, USA. Austin has a demographic makeup that school and 92% of the population has completed
includes 49%white population, 8% black or African secondary school. From six years, school
American population, and 8%Asian population. Austin attendance is compulsory, and 95% of children
contains a significant Hispanic or Latino ethnic minority aged six and over attend school. Austin has
(33.1%), with 11% of households speaking a language multiple public and private universities and
other than English, largely Spanish. Approximately half colleges, and approximately half of adults in
of the city’s population is religious, withmost religious Austin have bachelor’s degrees and 20% have
persons identifying as Christian. Austin is one of the postgraduate degrees. The national literacy rate
most educated andwealthy communities in Texas and is 99% of those 15 and over. There is a strong
the USmore broadly focus from early childhood on numeracy and
literacy. Schools are public, secular, monolingual,
co-ed, non-boarding, primary, and typically host
hundreds of students. The language of
instruction is English
Natal (Brazil) 895,000 Natal is located in the Northeast region of Brazil and is the 95% attend primary and 82% attend secondary
19thmost populous city in the country. 50% of school. There are tax-funded (public) and private
inhabitants identified as Pardo (variousmixed schools. School funding comes from all three
ancestries), 44% asWhite, 5% as Black, 1% as East Asian, levels of government. The national literacy rate is
and 0.1% as Natives or Indigenous. Brazilian Portuguese 93%. Nationally, Education is compulsory in
is the official language. Unemployment rate is, at 12%, Brazil between ages 7 and 14. Schools were
the sixth highest in the country (in 2022). Natal is located in Natal. They were public, secular,
representative of other urban areas locally and monolingual, co-ed, and not boarding. Though
throughout Brazil secular, Brazilian schools and businesses
commonly celebrate Christian dates. Primary
language of instructionwas Brazilian Portuguese.
Schools in Natal generally hold several hundred
pupils. Adults average 12 years of education
San Cristobal 404,350 San Cristobal is an area within the capital city of Colombia, 93% attend primary school, and 77% attend
(Colombia) Botaga, which itself has over 7million inhabitants. San secondary school, nationally. The national
Cristobal is a relatively poor area of the city. Many literacy rate is 96%, and national compulsory
families havemigrated into the area from rural, school begins at six years. Data was collected at a
conflict-ridden parts of Colombia within the last few public school in San Cristobal, inner Bogota. The
decades. Approximately 90% of the population are school has approximately 1,000 pupils, and the
Colombian, and 10%Venezuelan (recent immigrants). instruction language is Spanish. Most adults have
Spanish is widely spoken. San Cristobal is not particularly secondary level education, withminorities of only
representative of neighbouring communities, as it is near primary level (more in older generations) and
the centre of the biggest city in the country (Bogota), university level (more in younger generations)
leading to high proportions of internally displaced
people, recent immigrants, and commuters
(Continues)
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 of 17 RAWLINGS ET AL.
TABLE 2 (Continued)
Population
size (Approxi-
Population mate) Population description Educational information by population
Keningau 150,000 Keningau is situated in a valley surrounded bymountain 95% attend primary school, and 91% attend
(Malaysia) ranges and rain forests, situated about 65miles away secondary school. Schools of various sizes are
from the capital city of the state of Sabah (formerly widely available. Small private institutions are
known as North Borneo), EastMalaysia. It is the fifth also widely available to offer after-school or
biggest township of the state. Basic public holiday tutoring. Some schools operate in shifts
infrastructures such as health clinics, banks, restaurants, due to limited space. School funding comes from
and supermarkets are available. Keningau is ethnically the state, but policies are overseen by the school
and linguistically diverse. Residents predominantly district board.Most children start attending
speak eitherMalay orMandarin. Various temples, Kindergarten from 4 or 5 years of age, but all
churches, andmosques are available in the region. The childrenmust attend primary school by the age of
majority of indigenous people are Christian but also 7 years. The national literacy rate is 95%. The
practise traditional rituals. The study sample is schools that participants attended are public and
representative of peri-urban townships in the area, but secular, promotingmultilingual education.Most
not of bigger cities of them come from an indigenous, Christian
background. The primary language of instruction
is BahasaMalaysia (national language), however,
the schools also instruct in English orMandarin
Chinese, depending on subjects taught or the
theme of school activities. In the area, most
adults have high school level education
Blue 72,000 Participants were based in the BlueMountains region, 99% of children attend primary school, and 92%
Mountains which is a rural/suburban area approximately 100 attend secondary school. All childrenmust
(Australia) kilometres frommetropolitan Sydney, in the state of attend school between the ages of 6–16. Literacy
New SouthWales (NSW). It is less linguistically diverse and numeracy are top priorities for the
than Australia on average, and 90% of homes in the Blue Department of Education of NSW for all public
Mountains only speak English at home, while 67.6% is schools. The study school hosts 250 pupils, and is
the rate for NSWoverall, and 72% for Australia a public, secular, monolingual, co-ed, day-school.
The language of instruction is English. The
national literacy rate is 99% of those 15 and over
andmost adults have advanced diplomas or
bachelor’s degrees
Tanna 32,000 Participants were children based in three regions in Tanna, Around 78% of children attend primary school, and
(Vanuatu) Vanuatu. Tanna is a small island in the island archipelago 48% secondary school. Schools are partially
of Vanuatu. Some individual Ni-Vans still adopt a very funded by government and religious
traditional way of living in Kastom villages (Inkunala, and organizations, and still developing inmany areas.
Lenaualaul), while others adopt a less traditional lifestyle Around 65% of primary schools teach in English
and reside in themain town on the island (Lenekel). This and 35% teach in French. Currently, primary
community consists of urban and rural participants and education is not compulsory. The literacy rate is
is representative of other communities in the region 88% of those aged 15 and over. For the Tafea
Province, in which the community is located,
about 40% of adults have no school education,
39% have a primary school education, and 18%
have up to a secondary education. Less than 1%
of adults go on to tertiary or vocational school
Saltpond 25,000 Participants were Akan children from Saltpond, a small city Locally, 73% of children aged 6−11 years attend
(Ghana) that functions as the capital of theMfantsiman primary school and 57% attend secondary
Municipal District, in the Central Region of South Ghana. school. The literacy rate is 79%. Education is
TheMfantsemanMunicipal, with its Administrative compulsory from four years until 15 years of age.
Capital Saltpond, forms part of the 22Metropolitan The community contains several primary and
Municipalities andDistrict Assemblies (MMDAs) in the junior primary schools, as well as two secondary
Central Region and one of the 260MMDAs in Ghana. schools. Schools in Saltpond are public, and
Ethnically, 94% of the population is Akan and the methodist supported. The primary language of
primary language is the Akan language (or Fante). The instruction is English and Fante and pupils are
religious composition is distributed among Christians, predominantly Fante
Methodists, Protestants, andMuslims. This community is
representative of other local ones, though Saltpond has a
strong focus on petroleum extraction whereas other
communities have a greater focus on fishing industries
(Continues)
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAWLINGS ET AL. 7 of 17
TABLE 2 (Continued)
Population
size (Approxi-
Population mate) Population description Educational information by population
Motaba River 10,0001 Bandongo childrenwere recruited from a small multiethnic 65% of children in the Republic of the Congo attend
(Republic of village in the Likouala region of the Republic of the primary school and 18% attend secondary school.
the Congo) Congo, home to∼400 people. Bandongo people The national literacy rate is 77%. Education is not
primarily speak Lingala and Bondongo, however some enforceable where the research took place, so is
French phrases or nouns are occasionally used. Their dependent on family support. For those that
social structure exhibits strong age and sex hierarchies. attend, school in the area begins at 6 years. In the
The study village is very representative of other study population, there is one primary school, but
Bandongo communities attending secondary school requires relocating
to another village. The school is located on the
edge of a small village in the Likoulala region of
the Republic of the Congo. It is a public, secular,
co-ed, non-boarding primary school that is
primarily taught in French—despite French being
a seldom used language outside the classroom.
Most students were Bandongo, although precise
numbers are unavailable. In the study area, most
adults have not completed primary school
Waoroni 2,0002 Data was collected in Keweriono,Wentaro, Nenkipare, 91% of children attend primary school and 85%
(Ecuador) Tiguino and Bataboro communities, which are all attend secondary school, nationally. The national
relatively remote. All communities are primarily literacy rate is 94%. Schools are located in clear
ethnically and culturallyWaorani, and all also have some spaces, close to the forest. Schools in the sample
intermarriagewith Amazonian Kichwa groups. In area were public schools operating under the
multicultural families, theWaorani language (Wao intercultural and bilingual framework
Terrero) is typically dominant. Almost all individuals (undersecretary of education). This is one
under the age of 40 are fluent in Spanish, or a local modality that can be selected in public education
dialect of Spanish. Approximately half of the population as an alternative to the Hispanic education
identifies strongly as Christian (Evangelical or Catholic). system. These schools typically have teachers
Waorani society remains highly egalitarian in nature, and from indigenous nations in Ecuador. School sizes
high value is placed on individual autonomy. The range from 10–80 pupils, and the pupil
communities involved in this study very representative demographic is primarilyWaorani, and some
of otherWaorani communities in the region, however Amazonian Kichwa. Spanish is the primary
they are quite different from colonist communities, and language of instruction, althoughWaorani
are very different from urban centres teachers in primary school often teach inWao
Terrero or explain Spanish concepts in the local
language. In high school,Wao Terrero is
sometimes taught. All adults except elders over
the age of 80 years old have had some experience
with formal education.Most adults have had at
least three years of primary school, but very few
adults over the age of 45 have completed high
school
1Based on estimates from researchers and local residents.
2Based on ethnographic data (Cardoso et al., 2012).
3.2 School quality measures rience and composition, and variable 4 is related to the availability of
educational resources. Access to writing materials per classroom was
Researchers collecteddata on the followingmeasures of school quality, assessed by researchers at each site and reported on a scale from 0 to
either throughdirect observationor by interviewing teachers or school 3, where:
staffmembers. Ourmeasures of school qualitywere adapted frompre- 0=Nomaterials available
viously established measures which have, in some studies, previously 1=Very little available (i.e., fewer pencils, pens, paper than students
been shown to be important predictors of academic achievement and in the classroom)
performance (Burchinal et al., 2010; Davis et al., 2020; Mayer et al., 2= Some/adequate supplies available (i.e., just enough pencils, pens,
2000): (1) the number of students per classroom, (2) the number of paper for each student in the classroom)
grades or years per school, (3) the average number of teachers per 3 = Many supplies available (i.e., more than enough pencils, pens,
class, and (4) the amount of writing materials available to students paper for each student in the classroom; other writing materials such
(pens, pencils, notebooks). Variables1–3are related to classroomexpe- as markers, crayons, etc available to students in classrooms)
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 of 17 RAWLINGS ET AL.
3.3 Academic knowledge assessment ‘Rain falls down’, ‘Trees can grow very tall’). The experimenter uncov-
ered one sentence at a time and asked the participant to read the
The academic knowledge assessment task was a multi-stage assess- sentence aloud. For all sites, testing for the literacy section was ceased
ment of literacy and numeracy, and a similar version has been used if participants stopped reading on two consecutive sentences. For
with the Tsimané of Bolivia (Davis & Cashdan, 2019). It was based on the first four sites where data was collected (India, Congo, Vanuatu,
the Iowa Test of Basic Skills (Hoover et al., 2003), which has been used Ghana), it was also discontinued for children who made mistakes on
in multiple cultures (Clifford et al., 1989), but adapted for this project three consecutive sentences.
to focus on literacy and numeracy. Both the literacy and numeracy
components had four separate stages. The assessment was structured
such that if a participant failed to meet criterion of a given stage of 3.4.4 Reading comprehension
the numeracy or literacy components (see below), testing was ceased
at that point for that component. The order in which the literacy or The fourth stagewas reading comprehension. Participants read a short
numeracy components were presented was counterbalanced across one-paragraph story to themselves before answering four multiple
ages and gender, and the task was presented in the language that the choice questions about events and characters in the story.
participant was taught in school (though in populations where the pre-
dominant community dialect is different to the school language, the
instructions were given in the former). 3.5 Literacy component scoring
Scores were calculated for each of the four stages independently and
3.4 Literacy component overall. One point was given for a correct response on the letter and
word identification stages such that each had maximum scores of 10.
The academic knowledge assessment literacy component contained For the sentence reading, participants were given two points if a sen-
four subsections of increasing difficulty: letter identification, word tencewas readwith nomistakes, one point if one or twomistakeswere
identification, sentence reading, and reading comprehension. made on a sentence but the participant did not stop reading, and no
points if the participant made two or more mistakes and/or stopped
reading the sentence (maximum score of 20). Mistakes were classi-
3.4.1 Letter identification
fied as instances when the participant said the wrong word even if
In the first stage, letter identification, participants were presented they later corrected it, if they read the words out of order, and/or they
with a sheet of 14 randomly located different letters, with four black made an extreme mispronunciation. Stops were classified as instances
and white pictures spaced amongst them. The experimenter pointed when the participant stopped reading the sentence before it was com-
to one of these letters (out of 10 consecutively) and asked the par- pleted (i.e., the participant did not continue the sentence at any later
ticipant, ‘What letter is this?’ Participants were required to verbally point). Participants were given one point per correct answer on the
identify each letter the experimenter pointed at. If a participant reading comprehension stage (maximumscore of 4). Combining stages,
answered incorrectly for three consecutive problems within the letter participants could score amaximum of 44.
identification stage, testing for the literacy section was ceased.
3.6 Numeracy component
3.4.2 Word identification
The numeracy component also consisted of four subsections: a count-
The second stage of the literacy component was a word identification ing task, number identification, addition and subtraction mathematical
task. Participants were presented with a sheet of 14 randomly located problems, andmultiplication/division problems.
different but culturally accessible words (e.g., goat, sun, house) with
four black andwhite pictures spaced amongst them. The experimenter
pointed to one of these words (out of 10 consecutively) and asked 3.6.1 Counting task
the participant, ‘What word is this?’, and the participant was required
to identify this word. If a participant answered incorrectly for three The first stage of the numeracy component was a counting task. Par-
consecutive problemswithin the word identification stage, the literacy ticipants were presented with a sheet of different quantities of six
section was ceased. neutral stimuli images (i.e., one insect, two fish, three trees and so
on). The experimenter pointed to one of these images and asked the
3.4.3 Sentence reading participant, ‘How many [stimulus] is this?’. If a participant was incor-
rect on three consecutive problems within the counting stage, testing
The third stage was sentence reading. The experimenter presented for the numeracy section was ceased. There was a maximum of four
each participant with a list of ten sentences of increasing length (e.g., questions.
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAWLINGS ET AL. 9 of 17
3.6.2 Number identification TABLE 3 Descriptive statistics on key variables.
Total sample
The second stage was number identification, in which participants
Variable (units) Mean SD n
were presented with a sheet of 14 randomly located unique numbers.
Age (years) 8.69 2.23 889
The experimenter pointed to one of these numbers (out of 10 consec-
Schooling (years) 3.96 2.55 554
utively) and asked the participant, ‘What number is this?’ Participants
were required to identify this number. If a participant was incorrect Literacy component
for three consecutive problemswithin the number identification stage, Letter identification (0–10) 7.97 3.52 889
testing for the numeracy section was ceased. The maximum score was Word identification (0–10) 5.00 3.31 742
10. Sentence reading (0–20) 15.05 6.33 557
Reading comprehension (0–4) 2.79 1.19 484
Numeracy component
3.6.3 Addition/subtraction problems
Counting task (0–4) 3.67 0.83 884
Number identification (0–10) 9.00 2.38 855
The third stage consisted of addition/subtraction problems. The exper-
Addition/subtraction (0–6) 3.18 2.00 753
imenter presented the participant with a sheet of paper with six
addition/subtraction puzzles which varied in difficulty (i.e., 2 2 x, to Multiplication/division (0–5) 1.87 1.51 424+ =
72−37 = x) and said, ‘Okay, now let’s try these. Fill in the answers to School qualitymeasures
the problems on this sheet. Try as many as you can, even if you are Number of students per class 25.36 13.82 832
unsurewhat the right answer is. You can use this sheet to figure out the Number of grades per school 7.56 3.68 788
answers. Let me know when you’ve finished.’ Participants were given Number of teachers per class 1.28 0.41 889
up to six minutes to complete them and there was a maximum score of
Writing materials (0–3) 1.13 0.92 889
six.
3.6.4 Multiplication/division problems begin by describing the sample characteristics and descriptive statis-
tics of key variables by gender, controlling for age (Table 3). We then
The multiplication/division stage was the same as the addi- used linear models to assess performance on numeracy and literacy by
tion/subtraction, except children were presented with five age, controlling for gender. We next examined whether the different
multiplication/division puzzles (i.e., 35/7 = x, 29 × 53 = x) and measures of school quality are correlatedwith one another using Pear-
were given five minutes to complete them. If a participant was son’s partial correlations. Finally, we considered whether measures of
incorrect on four problems within the addition/subtraction or mul- school quality predict performance on measures of numeracy and lit-
tiplication/division stages, testing for the numeracy section was eracy using general linearmodels andmixed effectsmodels. To account
ceased. for the non-independence of learning institutions across populations,
eachpopulationwas entered as randomeffects intomixedeffectsmod-
els. Model comparisons provided estimates of the relative quality for
3.7 Numeracy component scoring statistical models, and best-fit models are reported in the results.
As with the literacy component, scores were calculated for each of the
four stages independently and overall. One point was scored for each 5 RESULTS
correct answer on the counting and number identification tasks, giv-
ingmaximum scores of 4 and 10, respectively. One pointwas also given 5.1 Literacy and numeracy
for each correct response for the addition/subtraction and multiplica-
tion/division problems, givingmaximum scores of 6 and 5, respectively. Figure 1 shows the distribution in performance on literacy tasks by
Combining each stage of the numeracy component, participants could gender, including letter identification, word identification, sentence
score amaximum of 25. reading, reading comprehension. Though boys showed slightly higher
averages on letter identification, there was no significant effect of
gender (β = −0.45, p = 0.06, 95% CI: −0.91, 0.02), but there was a sig-
4 STATISTICAL METHODS nificant positive age effect (β = 0.53, p < 0.001, 95% CI: 0.44, 0.63).
Boys did not outperform girls on any other literacy task (all ps > 0.05).
Data analysis and visualisation was conducted in R version 2022.02.0 However, therewerepositive associations betweenage andword iden-
using the packages ggplot2 for data visualisation (Wickham, 2009) and tification (β = 0.91, p < 0.001, 95% CI: 0.78, 1.04), sentence reading
lmer4 for model development and analysis (Bates et al., 2015). We (β = 1.25, p < 0.001, 95% CI: 1.00, 1.50), and reading comprehension
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
10 of 17 RAWLINGS ET AL.
F IGURE 1 Patterns in literacy scores. Density plots display distribution of boys’ and girls’ (A) letter identification (N= 889), (B) word
identification (N= 742), (C) sentence reading (N= 557), and (D) reading comprehension scores (N= 484). Sample sizes decreased as task difficulty
increased. Rugmarks on the X-axis indicate the range of average performance.
F IGURE 2 Patterns in numeracy scores. Density plots display distribution of boys’ and girls’ (A) counting (N= 884), (B) number identification
(N= 855), (C) addition and subtraction (N= 753), and (D) multiplication and division (N= 424). Sample sizes decreased as task difficulty increased.
Rugmarks on the X-axis indicate the range of average performance.
(β= 0.20, p< 0.001, 95% CI: 0.15, 0.26). Thus, literacy ability improved (β = −0.06, p = 0.31, 95% CI: −0.17, 0.05) and number identification
with age. (β = −0.07, p = 0.69, 95% CI: −0.41, 0.27) when controlling for age
Figure 2 displays the distribution of performance on the numer- (there were positive age effects; β = 0.13, p < 0.001, 95% CI: 0.11,
acy tasks. Boys and girls performed similarly on the counting task 0.16 and β = 0.37, p < 0.001, 95% CI: 0.29, 0.44, respectively). Older
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAWLINGS ET AL. 11 of 17
F IGURE 3 Patterns in literacy and numeracy scores across the sample. Scatterplots of literacy score (top row) and numeracy score (bottom
row) from 10 countries (N= 889). Both scatterplots are fitted with a smoothing spline for each sample of boys and girls. Shaded areas represent
95%CIs.
children performed better on the addition and subtraction (β = 0.55, TABLE 4 Pearson’s pairwise partial correlations between
p< 0.001, 95% CI: 0.49, 0.62) andmultiplication and division (β 0.43, measures of school quality (N= 889).=
p<0.001, 95%CI: 0.36, 0.50) tasks,when controlling for sex (β=−0.15, # Students # Grades # TeachersWriting
p = 0.26, 95% CI: −0.40, 0.11 and β = 0.03, p = 0.83, 95% CI: −0.24, per class per school per class materials
0.30, respectively). Figure 3 displays literacy and numeracy compos- # Students per class 1
ite scores (all sub-components combined) by age and gender, across # Grades per school −0.51o 1
populations.
# Teachers per class 0.13*** −0.18*** 1
Writingmaterials 0.26*** 0.36*** 0.21*** 1
6 RELATIONS BETWEEN MEASURES OF Statistical significance markers: o p ≤ 0.1; * p ≤ 0.05; ** p ≤ 0.01; *** p ≤
SCHOOL QUALITY 0.001.
Measures of school quality related to classroomcomposition and expe-
rience included the number of grades per school as well as the number We next examined the association between performance in liter-
of students and teachers per class. The amount of writing materials acy and numeracy with each measure of school quality, individually,
available to students reflected access to resources. Pearsonpartial cor- controlling for age and gender. Children at schools with fewer grades
relations between these variables are shown in Table 4. Controlling for demonstrated higher performance in literacy (β < 0.001, p = −6.43,
age and gender, the amount of writing materials available to students 95% CI: −7.01, −5.85) and numeracy (β = −2.16, p < 0.001, 95% CI:
positively correlated with the number of grades per school, and the −2.35, −1.97). Likewise, classrooms that had more teachers was also
number of teachers and students per class. The number of teachers per associated with higher literacy (β= 7.16, p< 0.001, 95% CI: 6.27, 8.05)
class and the number of grades per school were negatively correlated and numeracy (β = 2.64, p < 0.001, 95% CI: 2.35, 2.93). Greater avail-
with one another. The number of students per classwas negatively cor- ability of writing materials in the classroom also positively predicted
related with the number of grades per school and positively correlated children’s literacy and numeracy (β = 8.93, p < 0.001, 95% CI: 7.83,
with the number of teachers per class (see Table 4 for full correlations 10.04 and β = 3.43, p < 0.001, 95% CI: 3.07, 3.77, respectively). How-
between school quality variables). ever, there was no association between the number of students per
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
12 of 17 RAWLINGS ET AL.
F IGURE 4 Performance in literacy (left) and numeracy (right) as a function of school quality measures which were significant predictors,
controlling for age and gender. Performance in literacy and numeracy improves with increased access to writingmaterials. There was amoderate
positive effect of the number of teachers in the classroom on literacy performance.
classroom and literacy and numeracy performance (β = 0.01, p = 0.78, TABLE 5 Best-fit linear mixed effects and general linear models
95% CI: 0.06, 0.07). for literacy (left column) and numeracy (right column) by age, gender,−
andmeasures of school quality. Bothmixed effects models include
The three significant school quality predictors (teachers per class-
random intercepts for sample population.
room, grades per school, and writingmaterials) were then entered into
amixed effectsmodel. Eachmodel included the demographicmeasures Dependent variable: Performance on literacy
(age and gender) and school quality, with population entered in as a and numeracy
random effect (Figure 4). For the linear mixed models, which included (Mixed effects) (Mixed effects)
population as a randomeffect,models for literacy andnumeracy scores β β
(95%CI) (95%CI)
included age, gender, grades per school, writingmaterials, and teachers Literacy Numeracy
per classroom. Access to writing materials predicted greater perfor-
Age 3.89*** 1.50***
mance in both literacy and numeracy (Figure 5). There was a moderate (3.55, 4.23) (1.40, 1.61)
effect of the number of teachers per classroom on academic perfor-
Gender −1.0(−92.55, 0.38) -0.33
mance. The number of grades per school was positively associated but (Ref group: boys) (−0.77, 1.60)
did not significantly predict literacy or numeracywhen othermeasures Writingmaterials 6.37* 2.93**
of school quality were included in themodel (Table 5). (1.78, 4.76) (1.18, 4.67)
Teachers per class 0.93o 0.23
(0.01, 1.85) (−0.11, 0.58)
7 DISCUSSION Grades per school 7.93 1.66
(−0.03, 1.85) (−1.28, 4.61)
Recent decades have seen a marked increase in school attendance Constant -32.40 −4.96
across the world, driven by global initiatives aimed at providing all (−44.21,−20.59) (−9.33,−0.60)
children access to education. Yet, most of what we know about the Observations 785 778
impact of education on children’s learning is based on populations AIC 5937.11 4020.36
with multi-generational access to schools, funded and institutionalised
Statistical significancemarkers: o p≤ 0.1; * p≤ 0.05; ** p≤ 0.01; *** p≤ 0.0
at national levels. Research conducted in high-income countries has
shown that school quality has a marked, positive impact on the devel-
opment of academic skills such as numeracy and literacy. However, achievement remains underspecified. Here, we addressed these issues
we have little understanding of how to optimally measure school by investigating whether (1) there are age-related improvements in
quality and key predictors of academic achievement outside of high- academic achievement as evidence of general schooling effects, (2)
income populations. Moreover, there are several quantitative and school quality measures correlate with one another, and (3) they pre-
qualitative measures of school quality used in the literature, and the dicted children’s numeracy and literacy performance in 10 populations
validity of using specific measures to examine how it affects academic with diverse educational profiles. Our results show that improved
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAWLINGS ET AL. 13 of 17
F IGURE 5 Writingmaterials effects on literacy and numeracy scores. Box plots showingmean and first and third quartiles for literacy scores
(upper plot) andmathematics (lower plot).N= 889.
academic achievement is a product of the educational conditions in 7.2 Which measures of school quality were
which school-derived skills are taught and exercised, and that mea- associated with one another?
sures of school quality correlate with one another, within and across
domains. Across our school quality variables, measures of access to educational
resources (i.e., the amount of writing materials available to students)
and measures of within class composition (the number of grades per
7.1 Were there age-related and gender school and number of students and teachers per class) were positively
differences in academic achievement? correlated. Thus, larger schools, with larger class sizes andmore teach-
ers tended to be better resourced in terms of writing materials. We
Across populations, age was a strong, positive predictor of literacy and also found that variables related to classroom experience including the
numeracy for boys and girls. This provides evidence for general posi- number of teachers per class (which was a key predictor of academic
tive schooling effects on learning critical school-related skills. Research achievement) and the number of students per class were positively
on western populations has shown that as children get older and correlated, such that classes with more students had more teachers.
progress through schooling systems, their literacy and numeracy per- Conversely, the number of teachers and students per class were neg-
formance strongly improves (Aunio & Niemivirta, 2010; Weinberger, atively correlatedwith the number of grades per school, where schools
1996). However, other research in someGlobal South populations sug- with fewer grades hadmore teachers and students per classroom.
gests that the relationship between age and academic achievement is Although researchers have used a range of different measures of
complex and not as strong (e.g., South Africa: Spaull & Kotze, 2015), school quality, the validity of using diverse variables is not always clear,
potentially because varying school quality and wide wealth distribu- particularly because there is little information on whether variables
tion means many children acquire learning deficits early and fail to expected to be associated with one another are across populations.
catch up (Alcott&Rose, 2017; Spaull &Kotze, 2015). By demonstrating Whether measures of school quality can be adopted in populations
that, across multiple populations, with increased schooling exposure with diverse approaches to, and engagement with, formal education
their literacy and numeracy performance significantly improves, our has been the subject of recent debate, with some suggesting that
data suggests formal education is critical to the development of key cross-cultural approaches to assessing factors that predict scholas-
academic skills. tic success should be limited to ‘culturally-similar’ countries (Singer &
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
14 of 17 RAWLINGS ET AL.
Braun, 2018). An important component of reliability is that measures 7.4 Study implications
within domains are associatedwith one another, and this is particularly
important in global education research, given the range of potential Our data contribute to discourse on methods to develop optimal
measures of school quality available. Our finding that the variables educational environments, enabling children to reach their scholastic
assessed here were correlated with others in similar domains in 10 potential. In many remote communities—particularly some hunter-
culturally diverse countries contribute to this discussion, potentially gatherer, pastoralist, and subsistence agricultural populations—there
suggesting that they are valid indicators of school quality and can are many barriers precluding consistent access to formal schooling
be used as reliable measures to examine academic performance in for those who seek it (Ninkova et al., 2022; Siele et al., 2012). These
children and scholastic development. include economic, social, and cultural barriers, the stigmatisation chil-
dren face by peers at school, and disparities between the structured
and hierarchical approach of schools versus those of non-hierarchical
7.3 Which school quality measures predicted and transitional communities (Ninkova et al., 2022; Siele et al., 2012).
numeracy and literacy? Our data suggests that in multiple populations who follow diverse
educational approaches, access to educational resources and student-
We found that, across our study populations, more teachers per class- teacher ratios are crucial factors impacting core academic skills. With
roomwasmoderately, positively associatedwith literacy performance. widespread initiatives targeting increased access to formal schooling
Previous research, largely in western populations, has shown that across the globe, our findings emphasise, where possible, the need
reducing the student to teacher ratio in the classroom improves aca- to equip schools with basic resources and to consider the number of
demic success and provides population level health benefits (Adeyemi, teachers to maximise children’s academic achievements. In the face of
2007; Francis & Barnett, 2019; Harfitt, 2012; Muennig &Woolf, 2007; resource constraints, policymakers could consider alternative strate-
Pedder, 2006). Our findings showed that reducing the teacher to stu- gies to improve children’s learning experiences. For instance, some
dent ratio positively impacts literacy in diverse populations. This is research suggests that, in cases where reducing class size is unachiev-
ostensibly because reducing the student-to-teacher ratio allows chil- able, enabling teachers to work more intensively with small groups
dren tomore often, and directly, engagewith teachers and educational may provide some similar benefits to children’s academic achievement
resources and, conversely, reduces the load on teachers, leading to bet- (Sharples et al., 2019).
ter learning environments (Francis&Barnett, 2019). Previous research
in the west suggests that smaller classrooms may be more important
at younger ages, but that the relationship between class size and aca- 7.5 Study limitations
demic achievement is not straightforward and requires more research
given it is an expensive policy change (Ehrenberg et al., 2001; Jepson, A key limitation was that we cannot directly assess underlying mech-
2015).Our data indicates that population composition and subsistence anisms underpinning our findings. For example, because our data
levels also play a role.We thus argue that more qualitative data should were collected among school children, we could not address effects
be considered in global educational research and that a one-size-fits all of experiences outside of formal schooling on the development of
may not be optimal making policy decisions on class size. academic skills. Similarly, although we found that more teachers pos-
We also found that increased access towritingmaterials had strong, itively impacted academic performance, our dataset did not allow us
positive effects on children’s numeracy and literacy across populations. to examine why. Longitudinal research directly examining the underly-
Previous research in a range of countries in Southern Africa showed ing reasons that factors such as classroom size impact numeracy and
that greater access to writing materials was positively associated with literacy in diverse populations is an important next step for future
scholastic success because children can more readily engage in read- research.
ing and writing to practise and improve their academic skills (Hungi & Another limitation is that we have solely focussed on the relation-
Thuku, 2010). Here, we extend these findings to 10 globally and cultur- ship between school quality and performance on scholastic measures
ally diverse populations. Using pens and pencils to write also improves (numeracy and literacy). While this was our core aim, it does not allow
young children’s motor and visuospatial skill development—which assessment of other cognitive skills which may be positively or neg-
themselveshavebeen linked to academic performance—throughdraw- atively affected by formal education. It is important to consider, for
ing,writing numbers and letters and counting itemsonpaper (Cameron example, how teacher quality, or overall attendance to formal school-
et al., 2016). ing, may impact wider cognitive and social development in children
Conversely, the number of grades (years) per school did not signifi- in populations that have historically had little engagement with it.
cantly impact numeracy and literacy in our study, when controlling for For example, in many countries in the Global South, attending school
other measures of school quality. Given that most of our populations can contribute to local knowledge loss (by removing children from
hadsimilar knowncompulsory starting ages for school it is possible that core community activities) and disrupt language development (for
the grade ranges did not vary meaningfully across our sample. Further instance, when the language used in schools differs from children’s
research with populations who have larger variation in these variables mother tongues;Ninkova et al., 2022). Documentingwhether, and how,
is needed to verify these findings. attending school affects cognitive and social development is important
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAWLINGS ET AL. 15 of 17
to provide a holistic understanding of how formal education shapes the Institut National de Recherche en Sciences Exactes et Naturelles
children’s lives. (IRSEN), particularly Prof. Clobite Bouka-Biona (DG, IRSEN), for grant-
There are also trade-offs associated with different kinds of cross- ing us permission to conduct this research. From Vanuatu, we would
cultural comparisons. Far comparisons entail comparing populations like to thank Rachel Iaken, Annette Lalo, and Emma Ialimouk for sup-
that differ alongmany kinds of economic, social, cultural, linguistic, and port with fieldwork. We are thankful to the Vanuatu Cultural Centre
educational variables. The advantage of this kind of comparison is that for granting permission for this research andwe are especially thankful
it allows researchers to study variation that exists in some populations to Jean Pascal from the Tanna Cultural Centre for facilitating fieldwork
and not others. The disadvantage, however, is that the amount of vari- inKastom villages on Tanna and Lorine Sogari and the class teachers for
ability makes it difficult to isolate the impact of particular sources of permission to collect data in Harbour View primary school.
variation. Close comparisons entail comparing populations that are sim-
ilar according to most kinds of variables but differ in respect to a core CONFLICT OF INTEREST STATEMENT
variable of education (e.g., educational access). The advantage of close The authors declare no conflict of interests.
comparisons is that it allows researchers to test the impact of single
variables on key outcomes of interest. The disadvantage is that it only DATA AVAILABILITY STATEMENT
affords to study the range of variation that exists within largely similar Due to ethical reasons, this data is available from the corresponding
populations. In this study, however, we used both far and close cultural author upon reasonable request.
comparison, including high-, middle-, and low-income populations and
small-scale societies. This allowed us to sample awide range of globally ORCID
representative variation in school access and quality that would not be Bruce S. Rawlings https://orcid.org/0000-0001-9682-9216
possible by limiting our sampling to only far (or only close) comparison. LydiaChen https://orcid.org/0000-0002-4787-1597
NataliaDutra https://orcid.org/0000-0002-0766-0795
Frankie T. K. Fong https://orcid.org/0000-0002-6135-1379
8 CONCLUSION MicahGoldwater https://orcid.org/0000-0001-8052-9497
MarkNielsen https://orcid.org/0000-0002-0402-8372
Literacy and numeracy are increasingly considered fundamental skills JuliaWatzek https://orcid.org/0000-0002-9150-7469
of the 21st century, and, across the globe are associated with a range
of important life outcomes such as workplace achievement (Kuncel & REFERENCES
Hezlett, 2010) and social mobility (Mok & Neubauer, 2016). Our find- Adeyemi, T. O. (2007). The influence of class-size on the quality of output
ings have important implications for efforts by international agencies in secondary schools In Ekiti State, Nigeria. International Journal of Emo-
tional Psychology and Sport Ethics, 7(1). https://doi.org/10.4314/ijepse.
to improve numeracy and literacy in children. We find that age was
v7i1.38206
a strong predictor of academic achievement, suggesting that formal Alcott, B., & Rose, P. (2017). Learning in India’s primary schools: How
education has a marked impact on numeracy and literacy in popu- do disparities widen across the grades? International Journal of Educa-
lations with diverse educational profiles. We also find that several tional Development, 56, 42–51. https://doi.org/10.1016/j.ijedudev.2017.
05.002
measures of school quality are correlated in multiple populations, pro-
Aunio, P., & Niemivirta, M. (2010). Predicting children’s mathematical
viding key information on the validity of using different assays as performance in grade one by early numeracy. Learning and Individ-
reliable measures of the quality of schooling. Our data also reveal that ual Differences, 20(5), 427–435. https://doi.org/10.1016/j.lindif.2010.06.
more teachers in the classroom and increasing access to educational 003
materials such as books, pens, pencils, and notepads has a marked Ball, J., Paris, S. G., &Govinda, R. (2014). Literacy and numeracy skills among
children in developing countries. In Learning and education in developing
positive impact on children’s numeracy and literacy in populations countries (pp. 26–41). Palgrave Macmillan US. https://doi.org/10.1057/
varying in overall exposure to formal education, and their approaches 9781137455970_2
to schooling. Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-
effects models using lme4. Journal of Statistical Software, 67(1), 1–48.
https://doi.org/10.18637/jss.v067.i01
ACKNOWLEDGEMENTS
Burchinal, M., Vandergrift, N., Pianta, R., & Mashburn, A. (2010). Threshold
This research was supported by National Science Foundation grant analysis of association between child care quality and child outcomes
1730678 and Templeton Religion Trust grant TRT0206 to C.H.L. This for low-income children in pre-kindergarten programs. Early Childhood
research was also supported by grant P2CHD042849, awarded to the Research Quarterly, 25(2), 166–176. https://doi.org/10.1016/j.ecresq.
2009.10.004
Population ResearchCenter at TheUniversity of Texas at Austin by the
Burger,O., Chen, L., Erut,A., Fong, F. T.K., Rawlings, B., &Legare,C.H. (2022).
Eunice Kennedy Shriver National Institute of Child Health and Human Developing cross-cultural data infrastructures (CCDIs) for research in
Development.We thank all the children, caregivers, schools, communi- cognitive and behavioral sciences. Reviewof Philosophy and Psychology.
ties and research assistants for their help and collaboration throughout https://doi.org/10.1007/s13164-022-00635-z
Cameron, C. E., Cottone, E. A.,Murrah,W.M., &Grissmer,D.W. (2016). How
the project. From the Republic of the Congo, we would like to thank
are motor skills linked to children’s school performance and academic
Justin Ndambo for assistance in data collection, Adam H. Boyette achievement? Child Development Perspectives, 10(2), 93–98. https://doi.
for fieldwork support, Moïse Dzabatou for community support, and org/10.1111/cdep.12168
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
16 of 17 RAWLINGS ET AL.
Cardoso, S., Alfonso-Sánchez, M. A., Valverde, L., Sánchez, D., Zarrabeitia, Hungi, N., & Thuku, F.W. (2010). Differences in pupil achievement in Kenya:
M. T., Odriozola, A., Martínez-Jarreta, B., & de Pancorbo, M. M. (2012). Implications for policy and practice. International Journal of Educational
Genetic uniqueness of the Waorani tribe from the Ecuadorian Amazon. Development, 30(1), 33–43. https://doi.org/10.1016/j.ijedudev.2009.05.
Heredity, 108(6), 609–615. https://doi.org/10.1038/hdy.2011.131 001
Chasek, P. S., Wagner, L. M., Leone, F., Lebada, A., & Risse, N. (2016). Getting Imchen, A., & Ndem, F. (2020). Addressing the Learning Crisis: An Urgent
to 2030: Negotiating the post-2015 sustainable development agenda. Need to Better Finance Education for the Poorest Children.
Review of European, Comparative & International Environmental Law, 25(1), Jepson, C. (2015). Class size: Does it matter for student achieve-
5–14. https://doi.org/10.1111/reel.12149 ment? https://wol.iza.org/uploads/articles/190/pdfs/class-size-does-it-
Clifford, M. M., Lan, W. Y., Chou, F. C., & Qi, Y. (1989). Academic risk-taking. matter-for-student-achievement.pdf
The Journal of Experimental Education,57(4), 321–338. https://doi.org/10. Joynes, C., Rossignoli, S., & Amonoo-Kuofi, E. F. (2019). 21st Century Skills:
1080/00220973.1989.10806514 evidence of issues in definition, demand and delivery for development
Davis, H. E., Stieglitz, J., Kaplan, H., Tayo, A. M., & Gurven, M. (2020). contexts.
The formal schooling niche: Longitudinal evidence from Bolivian Ama- Kuncel, N. R., &Hezlett, S. A. (2010). Fact and fiction in cognitive ability test-
zon demonstrates that higher school quality augments differences in ing for admissions and hiring decisions. Current Directions in Psychological
children’s abstract reasoning. Preprint. https://doi.org/10.31234/osf.io/ Science, 19(6), 339–345. https://doi.org/10.1177/0963721410389459
d3sgq Legare, C. H., Dale, M. T., Kim, S. Y., & Deák, G. O. (2018). Cultural variation
Davis, Helen E., & Cashdan, E. (2019). Spatial cognition, navigation, and in cognitive flexibility reveals diversity in the development of executive
mobility among children in a forager-horticulturalist population, the Tsi- functions. Scientific Reports, 8(1). https://doi.org/10.1038/s41598-018-
mané of Bolivia. Cognitive Development, 52, 100800. https://doi.org/10. 34756-2
1016/j.cogdev.2019.100800 Legewie, J., & DiPrete, T. A. (2012). School context and the gender gap in
Davis, H. E., Stieglitz, J., Kaplan, H., & Gurven,M. (2021). School quality aug- educational achievement. American Sociological Review, 77(3), 463–485.
ments differences in children’s abstract reasoning, driving educational https://doi.org/10.1177/0003122412440802
inequalities. Evidence from a naturally occurring quasi-experiment in Lim, S. S., Updike, R. L., Kaldjian, A. S., Barber, R. M., Cowling, K., York, H.,
Amazonia, Bolivia. https://doi.org/10.31234/OSF.IO/D3SGQ Friedman, J., Xu, R., Whisnant, J. L., Taylor, H. J., Leever, A. T., Roman,
Dutra, N. B., Chen, L., Anum, A., Burger, O., Davis, H. E., Dzokoto, V. A., Y., Bryant, M. F., Dieleman, J., Gakidou, E., & Murray, C. J. L. (2018).
Fong, F. T. K., Ghelardi, S., Mendez, K., Messer, E. J. E., Newhouse, M., Measuring human capital: A systematic analysis of 195 countries and
Nielsen,M.G., Ramos,K., Rawlings, B., dosSantos, R.A.C., Silveira, L.G. S., territories, 1990–2016.The Lancet,392(10154), 1217–1234. https://doi.
Tucker-Drob, E. M., & Legare, C. H. (2022). Examining relations between org/10.1016/S0140-6736(18)31941-X
performanceonnon-verbal executive functionandverbal self-regulation Mayer, D. P., Mullens, J. E., Moore, M. T., & Ralph, J. H. (2000). Monitoring
tasks in demographically-diverse populations. Developmental Science. School Quality: An Indicators Report.
https://doi.org/10.1111/desc.13228 Mok, K. H., & Neubauer, D. (2016). Higher education governance in crisis:
Ehrenberg, R. G., Brewer, D. J., Gamoran, A., &Willms, J. D. (2001). Class size A critical reflection on the massification of higher education, graduate
and student achievement. Psychological Science in the Public Interest, 2(1), employment and social mobility. Journal of Education andWork, 29(1), 1–
1–30. https://doi.org/10.1111/1529-1006.003 12. https://doi.org/10.1080/13639080.2015.1049023
Else-Quest, N. M., Hyde, J. S., & Linn, M. C. (2010). Cross-national patterns Muennig, P., &Woolf, S. H. (2007).Health and economic benefits of reducing
of gender differences in mathematics: A meta-analysis. Psychological the number of students per classroom in US primary schools. Ameri-
Bulletin, 136(1), 103–127. https://doi.org/10.1037/a0018053 can Journal of PublicHealth,97(11), 2020–2027. https://doi.org/10.2105/
Enslin, P., & Horsthemke, K. (2016). Philosophy of education: Becoming AJPH.2006.105478
less western, more African? Journal of Philosophy of Education, 50(2), Ninkova, V., Hays, J., Lavi, N., Ali, A., da Silva Macedo, S. L., Davis, H. E., &
177–190. https://doi.org/10.1111/1467-9752.12199 Lew-Levy, S. (2022). Hunter-gatherer children at school: A view from the
Erbeli, F., Shi, Q., Campbell, A. R., Hart, S. A., & Woltering, S. (2021). Devel- Global South. Preprint. https://doi.org/10.31234/osf.io/zxq98
opmental dynamics between reading and math in elementary school. Nsamenang, A. B. (2005). Educational development and knowledge flow:
Developmental Science, 24(1). https://doi.org/10.1111/desc.13004 Local and global forces in human development in Africa.Higher Education
Francis, J., &Barnett,W. S. (2019). Relating preschool class size to classroom Policy, 18(3), 275–288. https://doi.org/10.1057/palgrave.hep.8300090
quality and student achievement. Early Childhood Research Quarterly, 49, Pedder, D. (2006). Are small classes better? Understanding relation-
49–58. https://doi.org/10.1016/j.ecresq.2019.05.002 ships between class size, classroom processes and pupils’ learning.
Gurven, M., Stieglitz, J., Trumble, B., Blackwell, A. D., Beheim, B., Davis, Oxford Review of Education, 32(2), 213–234. https://doi.org/10.1080/
H., Hooper, P., & Kaplan, H. (2017). The tsimane health and life history 03054980600645396
project: Integrating anthropology and biomedicine. Evolutionary Anthro- Pekkarinen, T. (2012). Gender differences in education. Nordic Economic
pology: Issues, News, and Reviews, 26(2), 54–73. https://doi.org/10.1002/ Policy Review, 1(1), 165–194.
evan.21515 Redding, C., & Nguyen, T. D. (2020). The relationship between school
Hanushek, E. A., & Woessmann, L. (2012). Schooling, educational achieve- turnaround and student outcomes: A meta-analysis. Educational Eval-
ment, and the Latin American growth puzzle. Journal of Development uation and Policy Analysis, 42(4), 493–519. https://doi.org/10.3102/
Economics, 99(2), 497–512. https://doi.org/10.1016/j.jdeveco.2012.06. 0162373720949513
004 Sepúlveda, D., Mendoza Horvitz, M., Joiko, S., & Ortiz Ruiz, F. (2022).
Harfitt, G. J. (2012). How class size reductionmediates secondary students’ Education and the production of inequalities across the Global South
learning: Hearing the pupil voice. Asia Pacific Education Review, 13(2), andNorth. Journal of Sociology, 58(3), 273–284. https://doi.org/10.1177/
299–310. https://doi.org/10.1007/s12564-011-9193-6 14407833211060059
Hoover, H. D., Dunbar, S. B., Frisbie, D. A., Oberley, K. R., Ordman, V. L., Sharples, J., Albers, B., & Fraser, S. (2019). Putting evidence to work: A
Naylor, R. J., & Shannon, G. P. (2003). Iowa test of basic skills guide to school’s guide to implementation. https://d2tic4wvo1iusb.cloudfront.
research and development. The Riverside Publishing Company. net/eef-guidance-reports/implementation/EEF_Implementation_
Humphreys, S., Moses, D., Kaibo, J., & Dunne, M. (2015). Counted in and Guidance_Report_2019.pdf?v=1675172629
being out: Fluctuations in primary school and classroom attendance in Siele, D., Swift, J., & Krätli, S. (2012). Reaching pastoralists with formal
northern Nigeria. International Journal of Educational Development, 44, education: A distance-learning strategy for Kenya. In Pastoralism and
134–143. https://doi.org/10.1016/j.ijedudev.2015.08.004 development in Africa (First, pp. 206–214). Routledge.
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAWLINGS ET AL. 17 of 17
Singer, J. D., & Braun, H. I. (2018). Testing international education assess- Weinberger, J. (1996). A longitudinal study of children’s early literacy expe-
ments. Science, 360(6384), 38–40. https://doi.org/10.1126/science. riences at home and later literacy development at home and school.
aar4952 Journal of Research in Reading, 19(1), 14–24. https://doi.org/10.1111/j.
Smithers, L. G., Sawyer, A. C. P., Chittleborough, C. R., Davies, N. M., Davey 1467-9817.1996.tb00083.x
Smith, G., & Lynch, J.W. (2018). A systematic reviewandmeta-analysis of Wickham, H. (2009). ggplot2: Elegant graphics for data analysis (Use R!).
effects of early life non-cognitive skills on academic, psychosocial, cog- World Bank. (2022). The State of Global Learning Poverty:
nitive and health outcomes. Nature Human Behaviour, 2(11), 867–880. 2022 Update. https://www.unicef.org/media/122921/file/
https://doi.org/10.1038/s41562-018-0461-x StateofLearningPoverty2022.pdf
Spaull, N., & Kotze, J. (2015). Starting behind and staying behind in
South Africa. International Journal of Educational Development, 41, 13–24.
https://doi.org/10.1016/j.ijedudev.2015.01.002
Tan, P. L. (2011). Towards a culturally sensitive and deeper understanding
of “rote learning” and memorisation of adult learners. Journal of Stud-
ies in International Education, 15(2), 124–145. https://doi.org/10.1177/
1028315309357940
Tobin, J.,Hsueh,Y.,&Karasawa,M. (2009).Preschool in three cultures revisited: How to cite this article: Rawlings, B., Davis, H. E., Anum, A.,
China, Japan, and the United States. University of Chicago PressChicago. Burger, O., Chen, L., Morales, J. C. C., Dutra, N., Dzabatou, A.,
Tout, D. (2020). Evolution of adult numeracy from quantitative literacy Dzokoto, V., Erut, A., Fong, F. T. K., Ghelardi, S., Goldwater, M.,
to numeracy: Lessons learned from international assessments. Interna- Ingram, G., Messer, E., Kingsford, J., Lew-Levy, S., Mendez, K.,
tional Review of Education, 66(2–3), 183–209. https://doi.org/10.1007/
Newhouse,M., . . . Legare, C. H. (2023). Quantifying quality: The
s11159-020-09831-4
UNESCO. (2011). International Standard Classification of Education. impact of measures of school quality on children’s academic
Voyer, D., & Voyer, S. D. (2014). Gender differences in scholastic achieve- achievement across diverse societies.Developmental Science,
ment:Ameta-analysis.Psychological Bulletin,140(4), 1174–1204. https:// e13434. https://doi.org/10.1111/desc.13434
doi.org/10.1037/a0036620
 14677687, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13434 by University of Ghana - Accra, Wiley Online Library on [08/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License