doesn’t have detailed information of family backgrounds. While this is available for grade 4

students, Jordan doesn’t participate in that version of TIMSS. In contrast, PISA data set includes

a wide range of indicators capturing household socio-economic status. The preferred socio-

economic status measure is the wealth index which is also used in the country-level descriptive

analysis.

Two sets of estimates are presented. First, using PISA data, OIC-wide analysis is undertaken. For

the sample of participating OIC countries as a group and contrasted with the same for the groups

of OECD and non-OECD non-OIC countries, to be presented in section 2 as part of the macro

analysis of education quality issues in the OIC. Second, country-specific regression analysis is

undertaken following the same approach in section 3 for Jordan and Malaysia as the underlying

data also comes from PISA 2012. The estimation strategy accounts for multiple plausible values

of the dependent variable.

In case of Nigeria and Pakistan, child level available assessment data corresponds to the primary

school level competency and come from two different sources which are not directly

comparable. Children tested also differ in terms of age group. Given differences in the sample

and underlying data set, it was not possible to maintain a fixed set of explanatory variables for

several reasons. Therefore, the full set of explanatory variables is not described here.

Nonetheless, certain variables have been included to ensure comparability (subject to

availability) in all country-specific analysis. These variables are described below.



Poverty: to describe poverty, the wealth quintiles generated by the authors have been

used.



School readiness: pre-school attendance



Other child-specific variable: the age and sex of the child, urban-rural residence, age and

sex of the household head.



Measure of intergenerational influence: Since none of the available data sets for study

countries have information on literacy outcomes for parents as well as children, it is not

possible to directly examine the extent of intergenerational transmission of illiteracy.

Nonetheless, it remains a serious issue in Nigeria and Pakistan where a large proportion

of children are first-generation learners and at-risk of remaining functionally illiterate

despite access to schooling. Therefore, in all cases, multivariate regression models at

least include parental schooling.

Given the stratifications in EGRA Nigeria survey, analysis of the raw data accounts use -svy-

command in STATA to account for the sample weighting. All regression models are estimated

using student final weight (i.e. wt_final) to scale to the population of males/females enrolled in

grades 2 and 3 for each State. Since students were tested in five subtests to measure

foundational to higher order literacy skills (letter sound identification, non-word coding, Oral

reading fluency (ORF), reading comprehension, listening comprehension) as part of the EGRA

assessment, multiple dependant variables are considered. The determinants of total scores are

However, sensitive check has been also performed using the index of Economic, Social and Cultural Status

(ESCS) constructed by the OECD. The index is constructed using information on a basket of 10 household items

that are common across participating countries: (i) a dishwasher; (ii) a DVD player; (iii) number of cellular

phones, televisions, computers, cars, rooms with a bath or shower; (iv) a room of their own; (iv) a computer that

can be used for schoolwork; (v) educational software; (vi) Internet; (vii) a desk; (viii) a quiet place to study; (ix)

books to help with school work and (x) reading materials and books. In addition, it includes three country

specific items. In order to document the extent of inequality in the level of student achievement, we use a number

of alternative proxy measures of household SES.