World Development Report 2018: Learning to Realize Education’s Promise

Measurement of learning shortfalls doesn’t provide clear guidance on how to remedy them. Fortunately, there is now a lot of experience on ways to improve learning outcomes at the student, classroom, and school levels. Cognitive neuroscience has evolved dramatically in the last two decades, providing insights on how children learn. This work has revealed how important the first several years of life are to a child’s brain development. At the same time, schools and systems around the world have innovated in many ways: by deploying novel approaches to pedagogy, using new technologies to enhance teaching and learning in classrooms, or increasing the accountability, and sometimes autonomy, of various actors in the system. The number of systematic evaluations of whether these interventions have improved learning has increased more than 10-fold, from just 19 in 2000 to 299 in 2016.

Many interventions have succeeded in improving learning outcomes. The learning gains from effective interventions translate into additional years of schooling, higher earnings, and lower poverty. For a group of stunted Jamaican children 9–24 months old, a program to improve cognitive and socioemotional development led to much better outcomes 20 years later—lower crime rates, better mental health, and earnings that were 25 percent higher than those of nonparticipants. Programs to improve pedagogy have had an impact greater than the equivalent of an extra half a year of business-as-usual schooling and an 8 percent increase in the present discounted value of lifetime earnings. So while tackling the learning crisis is hard, the fact that there are interventions that improve learning suggests ways forward.

This evidence base does not allow us to identify what works in all contexts because there are no global solutions in education. Improving learning in a particular setting will never be as simple as taking a successful program from one country or region and implementing it elsewhere. Randomized controlled trials and other approaches to evaluate impact place a premium on carefully isolating the causal impact of an intervention. But such approaches may ignore important interactions with underlying factors that affect whether an intervention makes a difference— factors that may not be at play when replicating the intervention in a new context. For example, increasing class size by 10 students reduced test scores by four times as much in Israel as it did in Kenya—and it has had no impact in some contexts. In the words of two commentators on this literature: “Knowing ‘what works’ in the sense of the treatment effect on the trial population is of limited value without understanding the political and institutional environment in which it is set.”

The next section tackles the question of that broader environment, but in the meantime we first address how to use this evidence most effectively. There are four main considerations.

First, more important than the individual results from individual studies are the principles of how and why programs work. In economic terms, “principles” correspond to models of behavior that can then help guide broader sets of approaches to addressing problems. Three types of models can prove especially insightful: straightforward models in which actors maximize their welfare subject to the constraints they face; principal-agent models that incorporate multiple actors with different goals and perhaps different information; and behavioral models that factor in mental models and social norms.

Second, a gap between what the evidence suggests may be effective and what is done in practice points to a potential entry point for action. Understanding why gaps open up helps guide how to address them. For example, when different actors face different information, or some actors lack information, this suggests drawing from approaches that show how information can be disseminated and used better. Gaps point to which types of principles should drive context-specific innovation.

Third, evidence tends to accumulate where it is easiest to generate, not necessarily where action would make the most difference, so policies focused only on that evidence might be misguided. Though the scope of the accumulated evidence in education is broad, just because an approach hasn’t been evaluated doesn’t mean it lacks potential. Context-specific innovation may mean trying things that have not been tried elsewhere.

Fourth, a focus on underlying principles highlights that the problem can’t be solved by one decision maker simply prescribing an increase in the quantity, or even the quality, of one or more inputs. Many of the inputs in learning are the result of choices made by the various actors—choices made in reaction to the actual and anticipated choices of other actors. For example, teachers respond to incentives to attend school and to improve student outcomes, even though the nature of the response varies across contexts. Likewise, students and parents make choices responding to other decisions. In India and Zambia, There are three key entry points to addressing learner government grants to schools led parents to reduce preparation:

their own investments in their children’s schooling. All things considered, a more complete character- • Set children on high-development trajectories through ization of the learning framework might be closer early childhood nutrition, stimulation, and care. Three to the one illustrated in figure O.14: learning how approaches stand out from successful experiences.

to improve outcomes by intervening at the student, First, target mothers and their babies with health classroom, and school levels involves illuminating the and nutrition interventions during the first 1,000 various arrows.

days to reduce malnutrition and foster physiologPutting all this together sheds light on three sets ical development. Second, increase the frequency of promising entry points: prepared learners, effec- and quality of stimulation and opportunities for tive teaching, and school-level interventions that learning at home (starting from birth) to improve actually affect the teaching and learning process.

language and motor development, as well as to

Each of these priority areas is founded on evidence cultivate early cognitive and socioemotional skills.

from multiple contexts showing that it can make a

Third, promote day-care centers for very young real difference for learning.

children and preschool programs for children 3–6 years old—along with caregiver programs that enhance the nurturing and protection of children—

Getting learners to school ready and motivated to to improve cognitive and socioemotional skills in learn is a first step to better learning. Without it, other the short run, as well as education and labor market policies and programs will have a minimal effect.

outcomes later in life. Program quality matters a lot: center-based programs with poor process quality (even with relatively good infrastructure, caregiver training, and caregiver-children ratios) can actually worsen developmental outcomes.

• Lower the cost of schooling to get children into school, but then use other tools to boost motivation and effort because cost-reducing interventions don’t usually lead to learning on their own. To improve learning, demand-side programs need to increase a student’s effort or capacity to learn. School-provided meals, for example, have had positive effects on access—and also on learning in places where children have limited access to food at home. Targeted cash transfers have led to more learning when they have incentivized performance itself or were marketed in a way that induces more effort, such as in Cambodia. Some information interventions have motivated efforts as well.

• To make up for the fact that so many youth lack skills when leaving basic education, provide remediation before further education and training. Remediation in school is a first best approach. After school, the more successful programs share two main features. First, they provide bridging courses in real-life settings, which allows learners with very low foundational skills to build these in the workplace. Second, accelerated, flexible pathways—not sequential courses over multiple semesters—are associated with greater student retention and ultimate certification.

Effective teaching depends on teachers’ skills and motivation, and yet many systems do not take them seriously. Teacher salaries are the largest single budget item in education systems, consuming three-quarters of the budget at the primary level in developing countries. Yet many systems struggle to attract strong candidates into teaching and to provide a solid foundation of subject or pedagogical knowledge before they start teaching. As a result, new teachers often find themselves in classrooms with little mastery of the content they are to teach. Once teachers are in place, the professional development they receive is often inconsistent and overly theoretical. In some countries, the cost of this training is enormous, reaching $2.5 billion a year in the United States. Moreover, education systems often have few effective mechanisms in place to mentor, support, and motivate teachers—even though teachers’ skills do nothing for learning unless teachers choose to apply them in the classroom. Fortunately, teachers’ skills and motivation can be strengthened, leading to greater effort and more learning, with three main promising principles emerging:

• For effective teacher training, design it to be individually targeted and repeated, with follow-up coaching—often around a specific pedagogical technique. This approach contrasts starkly with much of today’s professional development for teachers across a range of countries. In the United States, a team of teacher training experts characterized professional development there as “episodic, myopic, and often meaningless.” In Sub-Saharan Africa, teacher training is often too short to be effective and too low in quality to make a difference. By contrast, programs in Africa and South Asia that provided long-term coaching led to sizable learning gains.

• To keep learners from falling behind to the point where they cannot catch up, target teaching to the level of the student. Over the course of several grades, often only a fraction of learners progress at grade level, with most falling behind and some learning almost nothing. This is partly because teachers teach to the most advanced students in the class, as documented from Australia to Sweden to the United States, or because the curriculum is too ambitious but teachers are required to teach it. Effective strategies to target teaching to the level of the student include using community teachers to provide remedial lessons to the lowest performers, reorganizing classes by ability, or using technology to adapt lessons to individual student needs. • Use pecuniary and nonpecuniary incentives to improve the motivation of teachers, ensuring that the incentivized actions are within teachers’ capacity. Education systems typically neither reward teachers for performing well nor penalize them for performing poorly. Incentives are most likely to be effective at improving outcomes when there are straightforward actions that teachers can take to improve learning—such as increasing attendance when absenteeism is the constraint. But incentives do not need to be high stakes (or financial) to affect behavior. In Mexico and Punjab, Pakistan, simply providing diagnostic information to parents and schools about the schools’ relative performance improved learning outcomes.

School inputs, management, and governance must benefit the learner-teacher relationship if they are to improve learning—but many do not. Debates on improving education outcomes frequently revolve around increasing inputs, such as textbooks, technology, or school infrastructure. But too often the question of why these inputs might actually improve learning is overlooked. The evidence on successful use of inputs and management suggests three main principles:

• Provide additional inputs, including new technologies, in ways that complement rather than substitute for teachers. A computer-assisted learning program in Gujarat, India, improved learning when it added to teaching and learning time, especially for the poorest-performing students. A Kenyan program that provided public school teachers with tablets to support instruction increased the reading performance of their students. But simply providing desktop computers to classrooms in Colombia— where they were not well integrated with the curriculum—had no impact on learning. Even more traditional inputs—such as books—often fail to affect teaching and learning when they aren’t actually deployed in classrooms, or if the content is too advanced for the students. • Ensure that new information and communication technology is really implementable in the current systems. Interventions that incorporate information and communication technology have some of the biggest impacts on learning. But for every highly effective program—such as a dynamic computer-assisted learning program for secondary school students in Delhi that increased math and language scores more than the vast majority of other learning interventions tested in India or elsewhere—there are programs such as the One Laptop Per Child programs in Peru and Uruguay, which evaluations suggested had no impact on student reading or math ability. Technologies ill-adapted to their settings often fail to reach the classroom or to be used if they reach it. • Focus school management and governance reforms on improving teacher-learner interaction. Training principals in how to improve that interaction—by providing feedback to teachers on lesson plans, action plans to improve student performance, and classroom behavior—has led to a large impact on student learning. In countries ranging from Brazil and India to Sweden, the United Kingdom, and the United States, the management capacity of school principals significantly and robustly relates to student performance—even after controlling for a variety of student and school characteristics. Involving communities, parents, and school actors in ways that promote local oversight and accountability for service delivery can improve outcomes. But community monitoring tends to have more impact when it covers things that parents can easily observe (such as teacher absenteeism when it is high), and when a range of stakeholders (not just parents) are brought together in ways that lead to action. In Indonesia, school grants improved learning when links between the school and the village council—a center of local authority—were strengthened.

The most effective systems—in terms of learning—are those that have narrowed gaps between evidence and practice. On learner preparation, for example, East Asian countries such as Korea and Singapore have achieved high levels of children ready to learn. Stunting rates among preschool-age children are low, and children are motivated and supported by their families. To promote effective teaching, Finland and Singapore attract some of the most highly skilled graduates from tertiary education into teaching and provide them with effective professional development opportunities and sustained support.