Highlights We use nationally representative data to create mathematics learning profiles for Indonesia. We compare student learning levels and changes in learning from 2000 to 2014 to curriculum expectations. Students’ mastery of basic skills is low. Over 14 years, learning declined by approximately 0.25 standard deviations. The average child in grade 7 in 2014 had learned as much as the average child in grade 4 in 2000. Changes in learning were not driven by changes in student composition.

Indonesia has instituted wide-ranging educational reforms over the past twenty years, but recent international assessments of student learning indicate that these reforms may not have translated into learning gains-the country is performing comparatively poorly and worse than its regional neighbours. To examine the relationship between schooling completed and learning gains, and how that changed over time, we developed learning profiles using five rounds of data from the Indonesian Family Life Survey (IFLS). We show that Indonesia has succeeded in achieving high levels of school enrolment and attainment, with particular gains concentrated in junior secondary and senior secondary school between 2000 and 2014. However, we also find a large gap between students' mathematical ability and what they are supposed to know based on the education curriculum. Absolute learning levels as well as marginal learning levels are low, meaning that students are learning little as they are promoted from grade to grade. Even high school graduates struggle to correctly answer numeracy problems that they should have mastered in primary school. We also find that learning is decreasing slightly over time. We extend our analysis by identifying characteristics of children who are educationally left behind: children who are performing particularly poorly compared to their peers. Children with low numeracy levels are more likely to live in Eastern Indonesia, in rural areas, and be older and male. Our findings, albeit limited to a narrow set of test items, demonstrate the incredibly slow pace of learning occurring throughout Indonesia, and reiterate the importance of focusing system reforms on learning progress.

No abstract

A wave of teacher retirement in Indonesia provides an opportunity to replace them with better-performing teachers. We study whether teacher candidates' screening tests into Pendidikan Profesi Guru (PPG) or Teacher Professional Education, a postgraduate education programme in Indonesia, can predict their performance at the end of the programme and in an actual classroom situation at the beginning of their teaching career. Using administrative data of 1,291 primary school teacher candidates, we find that admission criteria, including undergraduate grade point average (GPA), online admission tests, and interview scores, can predict a candidate's performance on their knowledge and teaching practice exams at the end of their education programme. A one standard deviation higher online admission test score is associated with a 0.30 standard deviation higher score in the knowledge examination. Teacher candidates with a one standard deviation higher interview score perform 0.07 standard deviation better on the teaching practice examination. For teacher candidates with one standard deviation higher undergraduate GPA, their knowledge examination performance is 0.15–0.17 standard deviation higher on average, and their teaching practice exam score is 0.06-0.07 standard deviation higher on average. We then estimate the predictive ability of the admission criteria on student learning outcomes in numeracy and literacy, which uses 1,530 randomly sampled students taught by 114 teacher candidates. We find no evidence that the selection criteria predicted student learning in a meaningful way. Our results contribute to a nascent body of research on the selection of teachers using ex-ante criteria to identify effective teachers in developing countries.

Studies find that effective teachers raise student test achievement and lead to higher future earnings for the students (Chetty et.al, 2014; Hanushek, 2011). Teacher selection and the criteria used in making the selection are important because they aim to identify such effective teachers. Identifying teachers with such potential is relatively more cost-effective than other policies applied after the teachers have teaching jobs (Klassen and Kim, 2019; Hobson et al., 2010). Many studies focus on selecting teachers based on the information collected at the time of hire to predict student outcomes (Jacob et al., 2018; Hill et al., 2012; Staiger and Rockoff, 2010). Other studies identify potentially effective teachers even before they become teachers. Those studies use information from teacher education programme admission criteria to predict teacher candidates’ success in the programme (Heinz, 2013; Casey and Child, 2011; Caskey et al., 2001). Among teacher selection criteria, studies identified predictors of subsequent performance including undergraduate grades, written tests, interviews, and teaching practice. In developing countries, studies on teacher selection are virtually non-existent. We found two studies that focus on the selection of teachers during hiring. Both use candidates’ screening tests results to predict student learning outcomes (Araujo et al., 2020; Cruz-Aguayo et al., 2017). However, we did not find studies in developing country contexts that focus on selection of teachers into education programmes or how the admission criteria relate to student learning outcomes. Whether focusing on selecting teachers during their education programme or as they go through the recruitment process, studies on teacher selection across countries have the same underlying question: Will the criteria be able to identify effective teachers? The idea of teacher selection to improve the quality of the teaching force is appealing. For instance, in high performing countries in PISA, like Japan and Korea, where there are many teacher colleges (Ingersoll, 2007) and the most prevalent teacher employment is civil-service, great attention is paid to the quality of selection into teacher education programmes (OECD, 2018). Teacher selection is arguably more critical in developing countries. In most developing countries, the entry into teacher education programmes lacks selectivity and teacher qualifications tend to be set lower compared to other professional jobs (Béteille and Evans, 2019). Across all developing countries, a larger number of teachers are employed and account for most of the education spending, but their effect on student outcomes is small (ADB, 2021; Crawfurd and Pugatch, 2021). This suggests the need for more attention to policies such as the selection of teachers and criteria used to identify those best suited to teach in the classroom. In Indonesia, where the teacher recruitment system lacks a strong mechanism to ensure quality (Huang et al., 2020) and the teacher in-service training has not been effective (Revina et al., 2020), a potential way to improve the pool of teachers is through enhanced selection of individuals who will become teachers. We specifically question whether we can predict a teacher’s performance using information available when they were a teacher candidate. Admission criteria for teacher education are presumably intended to identify candidates who have the greatest likelihood of being able to do well in the academic programme and ultimately in the classroom as a professional. The identification of criteria that predict teacher subsequent performance would give policy makers a stronger understanding of where programme improvement may be needed.