Implications of learning trajectories for the global learning crisis

From Ed•sy•clopedia
Cultivating student learning

Learning levels vary a lot—and, often, are low.
Let's raise them.
Edsyclopedia icon cultivating student learning.png
Introductory articles
Key categories

One of the greatest development achievements of the late 20th and early 21st century has been the attainment of near universal primary school enrollment. Despite this demonstrable progress, many children who have been to school remain illiterate and innumerate. Low learning in developing countries has long been documented: the World Bank’s 1980 World Development Report on growth and human development highlighted that the quality of education “is generally low in developing countries” (World Bank, 1980, p52)[1]. Thirty-five years later, the 2015 World Education Forum called attention to the “alarming scale of the quality deficit in the global ‘learning crisis’” and set out “learning for all” goals to be achieved by 2030[2]. This “crisis” sentiment has been echoed by countless organizations since.

However, despite widespread acknowledgment of poor learning outcomes, many efforts and much dialogue today still focus on expanding access to schooling. This focus can be seen in the emphasis many actors place on expanding access to school, through means such as universal secondary and pre-primary school; on equality, despite low learning even among advantaged groups; and on “second-generation skills,” even while poor performance on basic skills such as literacy persists[3].

To date, the data needed to assess the dynamics of learning over age and grade have been sparse. This limited researchers’ and other stakeholders’ ability to weigh the benefits of expanding access versus improving learning, and left many questions relevant to conventional wisdom unanswered. Firstly, characterizing the trajectory of learning over the early grades is critical for priority setting. Early work on learning outcomes for example assumed that learning across grades was linear, but if learning becomes less steep with higher grades, particularly for those who are lagging, then even a low-learning linear slope is optimistic about likely improvements in learning from expanding schooling access. Furthermore, disaggregated slopes matter as much as the average slope, since learning for some groups may taper off more or less quickly than others. Secondly, while much attention is given to inequality, little is known about how early in the learning process inequalities emerge, and whether achieving equality across identifiable groups would do much to close gaps in learning outcomes within countries or between countries. Moreover, it is likely that the answers to these questions differ by country, and so analysis must be able to consider cross-country heterogeneity.

Existing data sources are often inadequate for addressing these issues. Established sources of learning data for developed countries, such as PISA and TIMSS, include few developing countries, and even for those that do participate, such assessments cannot address many critical questions. First, because they are administered only to select age or grade levels, they cannot track the trajectory of learning over multiple school years. Second, they are typically administered only to children who are in school, which is problematic for understanding learning among all children, as goals like universal literacy or numeracy require. This is particularly problematic in developing countries where completion rates are lower and dropout is higher, leaving more children out of such school-based exams. Among the seven developing countries that participated in PISA-D, for example, only 43 percent of 15-year-olds on average were eligible for the exam; in Cambodia only 28 percent were[4]. While more developing countries participate in regional assessments, such as SAQMEC, PASEC, and LLECE, these assessments suffer similar drawbacks to the international ones. National assessments often do not fill this gap, as they typically occur at the end of primary or secondary schooling and so do not offer information about learning across multiple grades nor about learning in the earlier schooling years, and some are not comparable over time.

Learning trajectories and policy implications

Recent papers on learning trajectories (also known as learning profiles) are helping to fill this gap. RISE-associated researchers have used data from more than 50 developing countries with more than six million respondents to develop learning profiles and provide detailed and nuanced insight on learning trajectories in developing countries. Learning trajectories illustrate the empirical relationship between years of schooling and learning achieved, showing how much children learn from one year to the next. Recent work analyzes how learning profiles differ across countries and across groups within countries. By using cohort-based learning assessments administered at the household level, such analysis provides unique insight on progress towards universal learning goals by providing information on children both in school and out of school. It sheds light on the tradeoffs inherent in goals aimed at achieving learning equity, defined as universal basic skills, and equality, defined as equalizing outcomes across identifiable groups such as girls and boys. Recent work has also simulated learning outcomes under a variety of scenarios. Collectively the work on learning profiles contributes to two main policy areas related to the prioritization and allocation of resources to improve learning in developing countries.

Learning gains from increasing grade attainment vs. steepening the learning trajectory

First, understanding and quantifying current learning trajectories helps to identify the relative benefits, in terms of improved learning, of increasing the years of schooling children attain versus improving children’s learning gains per year. There are tradeoffs in terms of resource allocation and prioritization between expanding years of schooling and improving learning, and many countries will struggle to make substantial progress on both at the same time under existing budgetary envelopes. Three recent papers contribute to this policy area using data from Demographic and Health Surveys, ASER and Uwezo surveys, and Financial Inclusion Insights surveys, respectively, to simulate counterfactual scenarios related to increasing years of schooling and improving learning based on observed learning profiles[5][6][7]. The simulations show how the tradeoffs differ across countries, and therefore how policy prescriptions should differ, based on current schooling and learning levels. For example, simulations using DHS data show that expanding years of schooling to universal primary completion in Ethiopia would improve women’s literacy by 57 percentage points, to 75 percent literate, and thus prioritizing more schooling could yield substantial gains[6]. In Nigeria, however, literacy would only increase by 5 percentage points under universal primary schooling, so prioritizing steepening the learning trajectory would yield much more learning[6].

Another paper contributing to this policy area simulates school dropout and examines how learning outcomes would change if children who drop out were instead required to complete more years of school. It shows that in contexts where dropout is at least in part driven by low learning, steepening the learning profile may do more to accomplish both attainment learning goals[8].

Two additional papers contribute to this policy area with in-depth examinations of specific countries. A study on Indonesia uses large scale panel data that tracked learning during a period of substantial reforms, including large increases in education spending and increased years of schooling attainment, and finds that learning during this period actually declined slightly, despite the substantial increased efforts. Possible explanations, and their implications for future policies, are discussed[9]. A paper on Pakistan uses unique panel data to show how learning trajectories differ by initial ability level. It finds that, while low performers are converging in ability with high performers between third and sixth grade, they were so far behind by third grade that they would need eight to ten additional years (more than remain in their schooling) to reach high performers’ levels[10]. The paper shows that in this context, expanding to secondary completion would still not achieve learning goals, and it highlights how critical learning in the early years is for determining learning trajectories in the later years.

Learning gains from equality, and its contribution to equity

The second policy area relates to how resources can be allocated globally and within countries to promote both equity (universal basic skills achievement) and equality (equalizing outcomes between groups). Analysis using disaggregated learning trajectory shows who low learners are, and whether low average learning outcomes are primarily driven by extremely low learning among specific groups (i.e. poor girls) or by low learning overall[7]. Collectively, the papers contributing to this area show that in many contexts low learning spans gender, income, and other distributions, and therefore broader reforms aimed at improving learning for all would achieve better outcomes than those aimed at education equality across certain groups[11].

Contributing to this area, a paper examining sources of inequality finds that ascriptive factors, such as gender, location, or socioeconomic status, account for a much smaller portion of inequality than what the authors term “pure inequality,” defined as the inter-quartile range of test-score distributions[12]. Their analysis finds that countries most often improve from low- to middle-performance (in international comparisons) by improving learning for the lowest achievers first. Also contributing to this policy area, additional papers analyze differences in schooling and learning between girls and boys, the poor and the better off, and other groups both within and across countries. Using learning profiles to simulate scenarios in which girls achieve the same schooling, learning, or both outcomes as boys, and in which the poor achieve the same schooling, learning, or both outcomes as the better off, these papers find that in many countries, achieving such equality would still leave many illiterate and innumerate, as learning is often low even among the more advantaged groups.

The need for systems level improvements

This literature, taken together, uses new large-scale, disaggregated learning data on cohorts of children across multiple grades in developing countries, to reinforce how devastatingly low learning is in many countries and how little children learn as they progress through school. Second, the papers show that where learning profiles are flat, investments in expanding the years of schooling will yield only small learning gains, and allocating resources between schooling expansion and learning improvements needs to be informed by context-specific learning trajectories. Third, simulations and analysis show that equality of schooling or learning across groups will often fall far short of equity goals of universal basic skills, since learning is low even among more advantaged groups. Achieving minimum learning standards for disadvantaged groups will often require much more than equality. This work suggests the need for improving learning at a systems level, such as through prioritizing basic skills in the early grades, aligning instruction with students’ skill levels, and reforming systems to be coherent for learning (for example, Banjeri & Chavan, 2016[13]; Muralidharan et al, 2019[14]; Piper et al, 2018[15]; and Pritchett, 2015[16]).

See also


  1. "World Development Report, 1980." World Bank.
  2. Lockheed, M., T. Prokic-Bruer and A. Shadrova (2015). “The Experience of Middle-Income Countries Participating in PISA 2000-2015.” PISA, World Bank, Washington, D.C./OECD Publishing, Paris.
  3. "GPE Impact." Global Partnership for Education.
  4. Schleicher, Andreas. (2018). PISA 2018 Insights and Interpretations. OECD.
  5. Akmal, Maryam & Lant Pritchett. (2019). Learning Equity Requires More than Equality: Learning Goals and Achievement Gaps between the Rich and the Poor in Five Developing Countries.
  6. 6.0 6.1 6.2 Pritchett, Lant, and Justin Sandefur. "Girls' Schooling and Women's Literacy: Schooling Targets Alone Won't Reach Learning Goals." University of Oxford and Center for Global Development.
  7. 7.0 7.1 Kaffenberger, Michelle, and Lant Pritchett. "How Close Would Gender Equality Get Us to Universal Literacy? Evidence from Learning Profiles from the Financial Inclusion Insights Data." University of Oxford.
  8. Kaffenberger, Michelle, and Lant Pritchett. "How Much More Would Children Learn If They Stayed in School? Simulating Endogenous Dropout." .
  9. Beatty, Amanda et al. (2018). "Indonesia Got Schooled: 15 Years of Rising Enrollment and Shallow Learning Profiles".
  10. Yi Chang, Andres & Jishnu Das. "Learning in Primary Schools: Evidence from a Large, Longitudinal Study in Pakistan". World Bank.
  11. Silberstein, Jason. (2021). "Measuring, Visualising, and Simulating Solutions to the Learning Crisis: New Evidence From Learning Profiles in 18 Countries".
  12. Crouch, L. and Rolleston, C. (2017). Raising the Floor on Learning Levels: Equitable Improvement Starts with the Tail.
  13. Banerji, Rukmini & Chavan, Madhav. (2016). Improving literacy and math instruction at scale in India’s primary schools: The case of Pratham’s Read India program. Journal of Educational Change.'s_primary_schools_The_case_of_Pratham's_Read_India_program
  14. Muralidharan, Karthik, Abhijeet Singh, and Alejandro J. Ganimian (2019). "Disrupting Education? Experimental Evidence on Technology-Aided Instruction in India." American Economic Review, 109 (4): 1426-60.
  15. Piper, Benjamin L., DeStefano, Joe, Kinyanjui, Esther, & Ong'ele, Salome A. (2018). “Scaling up successfully: Lessons from Kenya’s Tusome national literacy program.” Journal of Educational Change 19(3), 293-321.
  16. Pritchett, Lant (2015). “Creating Education Systems Coherent for Learning Outcomes: Making the Transition from Schooling to Learning,” RISE Working Paper #15.