Author : Martin Gustafsson
Site of the publication : media.unesco.org
Type of publication : Report
Date of publication : August 2025
Introduction
Despite some evidence of good educational practices in Africa and countries displaying improvements, on average Africa displays very low levels of learning proficiency. The commitment to addressing this problem, and learning from cases where success has been achieved, can be seen in for instance the African Union’s Nouakchott Declaration, which states that Africa should ‘end learning poverty by 2035’ in the early school grades. While this is an extremely ambitious aim, given how gradually educational improvement has occurred in even successful countries, improving learning outcomes substantially by 2035 is possible.
The continent has reached a critical point where between 79% and 97% of school-age children are in countries where there is at least some measure of learning from an international programme allowing for comparison across countries. The exact percentage depends on the definition of what can be considered sufficiently reliable data. The current level of coverage is a major achievement. However, for various historical reasons, often related to language and regional groupings on the continent, it has been difficult to compare statistics across different monitoring programmes.
Specifically, the paper aims to produce best possible estimates of the percentage of learners at the end of primary who have reached an adequate level of learning proficiency in reading and mathematics. Many important monitoring programmes have developed in the last couple of decades. This is done for 47 of 55 African countries, where these 47 countries cover 97% of Africa’s children aged 5 to 14. Adequacy with respect to learning outcomes is understood in terms of the Sustainable Development Goal (SDG) indicator 4.1.1b.
Stocktaking of existing statistics for Africa
The available statistics
Nine programmes are reflected in the table. The nine acronyms and associated full names are as follows:
- AMPL (Assessment for Minimum Proficiency Level);
- EGRA (Early Grade Reading Assessment);
- LaNA (Literacy and Numeracy Assessment);
- MICS (Multiple Indicator Cluster Surveys); PAL (People’s Action for Learning);
- PASEC (Programme d’analyse des systèmes éducatifs de la Confemen);
- PIRLS (Progress in International Reading Literacy Study);
- SACMEQ (Southern and Eastern Africa Consortium for Monitoring Educational Quality);
- TIMSS (Trends in International Mathematics and Science Study).
What is not within the scope of the current paper is data collected at the secondary level, such as PISA for Development (PISA-D). In Africa such data are less available than for the primary level and would not have covered countries which were not already covered by the primary level data.
The continent has reached a critical point where between 79% and 97% of school-age children are in countries where there is at least some measure of learning from an international programme allowing for comparison across countries. The exact percentage depends on the definition of what can be considered sufficiently reliable data. The current level of coverage is a major achievement. However, for various historical reasons, often related to language and regional groupings on the continent, it has been difficult to compare statistics across different monitoring programmes
To illustrate the slowness of improvement in learning outcomes, a phenomenon which makes it feasible to use relatively old statistics for some countries. PASEC 2014 and 2019 proficiency 15 statistics can be compared. Across the ten PASEC countries with results in both years, the gain in the percentage of learners who were proficient in mathematics at the end of primary was minus 0.5 percentage points a year .
Of the 145 national statistics appearing, 91% are from 2013 or later. And 94% of the 47 countries with statistics have a statistic that is not older than 2013.
Characteristics of the nine programmes
Roughly in order from greater to lesser ability to produce the desired proficiency statistics, the programmes are:
- Trends in International Mathematics and Science Study (TIMSS); TIMSS is a longstanding programme of the IEA8 with excellent documentation and microdata that are easily accessed. TIMSS has been used to define a minimum level of proficiency in mathematics for SDG indicator 4.1.1b, which deals with the end of primary.A 2021 UNESCO guide stipulates that what TIMSS defines as the ‘intermediate benchmark’ should be considered the minimum standard for mathematics at the end of primary. The advantage with TIMSS (and PIRLS) is that the long history of the programme has allowed lessons to be learnt, and that the transparency of the programme instils trust. A disadvantage for the current work is that the footprint of TIMSS is limited within Africa.
- Progress in International Reading Literacy Study (PIRLS); The advantages and disadvantages of PIRLS for the current work are essentially the same as for TIMSS.
- Literacy and Numeracy Assessment (LaNA);
- Assessment for Minimum Proficiency Level (AMPL) ; One clear advantage for the current analysis with AMPL is that it is the only programme used here which was specifically designed to gauge proficiency in terms of minimum competencies identified by UNESCO for the purposes of monitoring SDG goal 4.
- Programme d’analyse des systèmes éducatifs de la Confemen (English: Program for the Analysis of Educational Systems of CONFEMEN) (PASEC); A key advantage with PASEC is that it covers many African countries, that its level of difficulty is based on what skills learners in Africa typically display24, and that the level of transparency, in terms of technical documentation and microdata availability, is relatively good. The Africaonly focus would however limit comparisons beyond the continent.
- Southern and Eastern Africa Consortium for Monitoring Educational Quality (SACMEQ) ; MICS; MICS is a major and well-documented UNICEF programme collecting data from nationally representative samples of households, with a focus on children’s health and well-being. A disadvantage with MICS is its ‘ceiling effects’, meaning its tests are set at a relatively low level of difficulty, thus precluding meaningful differentiation of proficient and non-proficient learners in grade 6.
- People’s Action for Learning (PAL) ; PAL is another household-focussed programme with a learning assessment, focussing on a small group of countries.
- Early Grade Reading Assessment (EGRA). Finally, EGRA involves assessing children in schools on reading. Its limitations for the current study include that its focus on across-country comparability is not strong, and that its microdata are not readily available.
The gain in the percentage of learners who were proficient in mathematics at the end of primary was minus 0.5 percentage points a year
Harmonising national statistics
The overall approach taken
It is instructive to examine the similarities and differences between the hybrid harmonisation process presented in the current paper, and typical harmonisation exercises in the past. Gust et al (2024) can be considered representative of the latter. That is also the most recent major work of its kind, and produced what is easily the most comprehensive set of harmonised learning measures for the world. Gust et al produced proficiency statistics for 159 countries relating to mathematics at the lower secondary level. While 85% of the world’s population is covered by the assessment data used, this drops to just 50% in the case of Africa, where assessment data from 29 of 55 countries were available.
The hybrid harmonisation presented below responds to a somewhat different need to that of, for instance, Gust et al. The latter is strongly focussed on producing a global aggregate statistic, and global patterns that can inform analysis of, for instance, the impact of changing levels of hegemony across the world on subsequent economic growth. The aim below is more to inform discussions in a region, namely Africa, around the distribution and depth of ‘learning poverty’, how to resolve this, and also how to improve the monitoring of the situation.
To a significant degree, the audience is assumed to be national researchers and policymakers in the education field with a specific interest in their country. This explains why the emphasis is on drawing from whatever proficiency statistics exist, even if the underlying microdata are not accessible.
Informing grade-on-grade gain assumptions
The current paper assumes that grade 6 should be compared across countries, even if the SDG indicator in question refers to ‘end of primary’. It is thus assumed that it is unfair to compare, say, Zambia’s grade 7 proficiency to Côte d’Ivoire’s grade 6 proficiency, even if in both cases this represents the end of primary schooling. It would be unfair because Zambia’s learners have enjoyed the benefit of one extra year of schooling.
To illustrate a result, If the initial level of proficiency P1 is 20%, if the gain in standard deviations is 0.30, and an adjustment of just one grade is needed, then the new level of proficiency P2 is 29.4%. The normal distribution moves to the right by 30 points, which is 0.30 of the standard deviation of 100.
Step 2: Removal of implausible values
In this step the aim was to identify country values which emerged as outliers when programmes were compared, with a view to excluding such country values, especially if either the country was not a linking a country, or if a country value had apparently a more reliable alternative elsewhere in the data.
Selection of best statistics per country and a reality check
The figures, when combined with child population figures, allow for an Africawide aggregate proficiency statistic to be calculated. The across-subject average for the continent is 13.4%, considering countries with 97% of the continent’s children. The subject specific proficiency values are 16.1% for reading and 12.5% for mathematics, though here only 79% of population-weighted Africa is considered.
The 47 countries referred to account for 97% of Africa’s children. As discussed in section 3.1 above, Angrist et al (2021) also present harmonised values, though they are mean scores and not proficiency statistics, for 94% of the continent.
How considering the out-of-school changes the picture
The definitions of the SDG 4.1.1 indicators, and actual SDG-related statistics, exclude children not in school. Yet there is inevitably much interest in estimating proficiency levels in the young population as whole, regardless of school participation.
The proficiency values represented produce an aggregate of 13.4% when child populations are used as weights. This drops to 10.8% if out-of-school adjustments are taken into account. Put differently, only 11% of young Africans attain the minimum proficiency level for end of primary as understood within the SDG monitoring system.
How lower primary predicts end of primary
The current report has focussed on the harmonisation of proficiency statistics at the end of the primary level. In many countries there are high numbers of children aged well above age 7 in grade 1 – for instance in Guinea Bissau (GNB) 69% of grade 1 learners are older than 7, against just 9% in Tunisia. Moreover, by age 14 a high proportion of learners are still in the first three grades – this is true for a quarter of age 14 learners in Madagascar.
To what extent are the differences in grade 1 across countries due to children entering school with better capabilities because of factors such as nutrition and pre-school educational activities, at home or in a pre-school? Much of the proficiency difference across countries among enrolled learners in the early grades is due to what is learnt right at the start of schooling.
The proficiency values represented produce an aggregate of 13.4% when child populations are used as weights. This drops to 10.8% if out-of-school adjustments are taken into account. Put differently, only 11% of young Africans attain the minimum proficiency level for end of primary as understood within the SDG monitoring system
Countries that display relatively high levels of proficiency at the end of primary also do so in the early grades. Differences across countries in the proficiency of learners in grade 1, or children aged 7, are very large, and on the whole larger than the gains children make within individual countries. This raises important questions around the extent to which factors such as nutrition in the very early years play a role, as opposed to a ‘burst’ of skills generation in the very first year of so of schooling. Evidence-based answers to these questions in the African context remain scarce.
The challenge of measuring improvement over time
The SDGs focus primarily on improvement within countries, and not directly on across-country comparability. It is very noteworthy that despite apparently remarkable improvements seen in SACMEQ and to some extent PASEC, this has received so little attention in government documents in the affected countries. Government plans and reviews across the continent, and possibly beyond, often fail to focus on progress in terms of comparable measures of learning outcomes.
Conversations with education planners moreover suggest that the disappointing nature of the proficiency statistics in terms of levels of learning easily detract from the importance of improvements. Put differently, even relatively good improvements appear too slow to be interesting for governments.
Raising the status of programmes such as PASEC and SACMEQ would improve the possibility that gains in learning outcomes in African countries are recognised, not just within the continent but also worldwide. Raising their status would mean ongoing investments to ensure these programmes enjoy skilled analysts and implementers, but also investment in high quality technical products, particularly metadata and microdata. It is worth reminding stakeholders on the continent that not just having educational improvements, but also having indisputable evidence of this, improves the attractiveness of countries for local and international investors, for whom a skilled workforce is important.
Conclusion and pointers for future work
The current paper produces across-subject average measures of learning proficiency for countries representing 97% of the continent’s children. In what has been presented in the current paper, 26 of African 47 countries draw from data which are from the years 2018 to 2023, representing a significant refreshing of the data. This is made possible largely due to the 2019 PASEC results and data having been disseminated, and the availability of MICS foundational learning data, LaNA and AMPL, three sources not available before 2018.
Countries that display relatively high levels of proficiency at the end of primary also do so in the early grades. Differences across countries in the proficiency of learners in grade 1, or children aged 7, are very large, and on the whole larger than the gains children make within individual countries. This raises important questions around the extent to which factors such as nutrition in the very early years play a role, as opposed to a ‘burst’ of skills generation in the very first year of so of schooling. Evidence-based answers to these questions in the African context remain scarce
Firstly, it seems important to let the analysis be informed by a clear understanding of the users of the final statistics. Secondly, harmonisation exercises provide an opportunity to explore why certain countries display better proficiency, or better progress over time. Thirdly, an overly narrow view of what data are sufficiently reliable to warrant analysis and what a sufficiently rigorous harmonisation methodology looks like should be avoided. Fourthly, whether measures of learning should be imputed using non-assessment data is a question that is often asked. Fifthly, the role of national, as opposed to international, assessments is an important matter.
Turning to what the paper suggests in relation to the utility of existing systems, what stands out is the problem of microdata not being widely available to researchers. A likely reason for the non-availability of microdata is that programmes lack the resources to prepare the data and accompanying manuals for public release. This resourcing problem should be addressed. However, discussions with people involved in the process also suggest that there is sometimes a perception that making the data available increases the risk that errors in the published statistics will be exposed, leading to harm to the assessment programme’s reputation.
While there is some truth to this, the reputational damage caused by a lack of transparency is likely to be far greater. Conversely, quality control by a wide range of users of the microdata, including researchers not employed by the programme, helps to enhance a programme’s status.