letters from the alphabet, ‘letter’ level if they can successfully identify letters from the alphabet,

‘word’ level if they can identify words, ‘sentence’ level if they can read a sentence fluently and

‘story’ level if they can successfully read a story. For the purposes of the analysis, we re-coded

the achievement level scores to mean: (1) The child was coded at ‘higher’ language level (equals

1) if he or she could read a story (in Urdu, Sindhi or Pashto), and 0 if not; (2) The child was coded

at ‘higher’ mathematics level (equals 1) if he or she reported at ‘division’ level, and 0 if not. Probit

models were estimated with ‘higher’ levels in language and mathematics as dependent variables

and various covariates as independent variables. This is done for data from 2013 and 2016, two

rounds with full district coverage and a three year period between them to identify any

measurable changes over time.

In case of Nigeria, a number of providers cater to educational demand in Nigeria. Alongside non-

religious government schools, there are non-religious private schools as well as thousands of

Islamic schools or madrasahs. While the latter operates in different forms, the three main types

of Islamic schools in Nigeria are Qur’anic, Islamiyya, and Tsangaya. These madrasahs operate at

all levels of school education -- kotso (nursery stage), tittibiri (elementary stage), k’olo (middle-

level) and, culminates (higher level). Almost four in every five Muslim children attend at least

one type madrasah. IQTE (non-formal integrated Qur’anic/Islamiyya and Tsangaya Education)

offers a consolidated form of Islamic education. However, data on the quality of this type of

madrasah along with comparable information on the performance of government school

students is rare. Access to government data on national examination performance is also not

publicly accessible. Therefore, the data used comes from USAID’s EGRA-SSME cross-sectional

study of grades 2 and 3 students conducted in 4 states of Nigeria -- Jigawa, Kaduna, Kano, and

Katsina. Two-grade sample enabled an assessment of any improvement owing to an extra year

spent in school. The focus on two languages allowed an assessment of proficiency in the two

official languages that are used as the medium of instruction for the basic education curriculum

in both formal and non-formal schools. Given the focus on grades 2 and 3, schools which did not

have these grades were excluded from the study sample. Data was collected based on a two-

stage sampling methodology. In the first stage, 258-Schools stratified by school type were

sampled; 31 schools were selected per type in each state, proportional to grade 2 and grade 3

enrollment. In the second, stage, 3,795-Students were selected, stratified by grade and gender

where 5 students from each gender in each grade were sampled with equal probability.

Stakeholders’ Interviews and Review of Secondary Literature

To complement the quantitative analysis, a number of stakeholders’ interviews have been

carried out in each country. Stakeholder’s beliefs about the current problems in the educations

sector in their country matter because it influence their attitudes toward and support for reform

initiatives address these problems. But these beliefs may not be necessarily accurate. Moreover,

views of policy stakeholders in centralized system may not be aligned with those directly

involved with day to day management of the school.

While the exact number of interviews competed varies from country to country, at least 12

stakeholders were successfully interviewed in each case study sample country. The same

interview protocol was used in each country, although some questions were reworded to better

reflect local realities. In addition, country-specific questions were also included. The interviews

were meant to capture the perceived barriers to the delivery of quality education as well as

possible policy solutions. Depending on the expertise of the respondent stakeholder, however,

discussion also covered additional country-specific issues. In terms of field management, the

interviews were face to face in most cases. If the respondent did not speak English, a local