ľ¹ÏÓ°Ôº

All children should be able to read by age 10. Reading is a gateway for learning as the child progresses through school¡ªand conversely, an inability to read slams that gate shut. Beyond this, when children cannot read, it¡¯s usually a clear indication that school systems aren¡¯t well organized to help children learn in other areas such as math, science, and the humanities. And although it is possible to learn later in life with enough effort, children who don¡¯t read by age 10¡ªor at the latest, by the end of primary school¡ªusually fail to master reading later in their schooling career.

In recent years, it has become clear that many children around the world are not learning to read proficiently. Even though most children are in school, a large proportion are not acquiring fundamental skills. Moreover, 260 million children are not even in school. This is the leading edge of a learning crisis that threatens c and achieve the Sustainable Development Goals (SDGs). Without foundational learning, students often fail to thrive later in school or when they join the workforce. They don¡¯t acquire the human capital they need to power their careers and economies once they leave school, or the skills that will help them become engaged citizens and nurture healthy, prosperous families. As a major contributor to human capital deficits, the learning crisis undermines sustainable growth and poverty reduction.  

To spotlight this crisis, we are introducing the concept of Learning Poverty, drawing on new data developed in coordination with the UNESCO Institute for Statistics.  Learning poverty means being unable to read and understand a simple text by age 10. This indicator brings together schooling and learning indicators:  it begins with the share of children who haven¡¯t achieved minimum reading proficiency (as measured in schools) and is adjusted by the proportion of children who are out of school (and are assumed not able to read proficiently).

Using a measure developed jointly by the World Bank and UNESCO¡¯s Institute of Statistics, we have determined that 53 percent of children in low- and middle-income countries cannot read and understand a simple story by the end of primary school. In poor countries, the level is as high as 80 percent. Such high levels of illiteracy are an early warning sign that all global educational goals and other related sustainable development goals are in jeopardy.

Progress in reducing learning poverty is far too slow to meet the SDG aspirations:  at the current rate of improvement, in 2030 about 43% of children will still be learning-poor. Even if countries reduce their learning poverty at the fastest rates we have seen so far in this century, the goal of ending it will not be attained by 2030.

There is an urgent need for a society-wide commitment to invest more and better in people. If children cannot read, all education SDGs are at risk. Eliminating learning poverty is as important as eliminating extreme monetary poverty, stunting, or hunger. To achieve it in the foreseeable future requires far more rapid progress at scale than we have yet seen.

The learning poverty indicator focuses on reading for three reasons:

1. Reading proficiency is an easily understood learning measure
2. Reading is a student¡¯s gateway to learning in other areas
3. Reading proficiency can serve as a proxy for foundational learning in other subjects

The learning poverty indicator allows us to illustrate progress toward SDG 4¡¯s broader goal to ensure inclusive and equitable quality education for all. It particularly highlights progress towards SDG 4.1.1(b), which specifies that all children at the end of primary reach at least a minimum proficiency level in reading.

Methodology

The indicator combines the share of primary-aged children out-of-school who are schooling deprived (SD), and the share of pupils below a minimum proficiency in reading, who are learning deprived (LD). By combining schooling and learning, the indicator brings into focus both ¡°more schooling¡±, which by itself serves a variety of critical functions, as well as ¡°better learning¡± which is important to ensure that time spent in school translates into acquisition of skills and capabilities. 

How Learning Poverty is defined

The learning poverty indicator is calculated as follows:

LP = [LD x (1-SD)] + [1 x SD]

LP = Learning poverty

LD = Learning deprivation, defined as share of children at the end of primary who read at below the minimum proficiency level, as defined by the in the context of the SDG 4.1.1 monitoring

SD = Schooling deprivation, defined as the share of primary aged children who are out-of-school. All out-of-school children are assumed to be below the minimum proficiency level in reading.

Learning poverty can be improved in two ways: (i) by reducing learning deprivation as countries raise proficiency levels for children below the minimum proficiency threshold, or (ii) by reducing schooling deprivation as countries expand coverage and bringing out-of-school population into the system.

While schooling deprivation can be directly observed depending on whether the child is enrolled or not enrolled in school, learning deprivation cannot be directly observed, and is measured through standardized assessments using SDG¡¯s definition of minimum proficiency level, where reading proficiency is defined as:

Three complementary concepts: Learning poverty level, gap, and severity

The learning poverty level (or headcount ratio) shown above, that is the share of 10-year-olds who are not in school (schooling deprived) or are below the minimum proficiency level (learning deprived), has limitations. It does not capture the average learning shortfall among children under the minimum proficiency level. Hence, we include the learning poverty gap, that measures the average distance of a learning deprived child to the minimum proficiency level and indicates the average increase in learning required to eliminate learning poverty.

However, the gap measure cannot distinguish between an increase in the learning gap driven by students near the threshold and one driven by those at the very bottom of the learning distribution. Learning poverty severity captures the inequality of learning among the learning poor population and is the gap squared in relation to the minimum proficiency squared.

The concepts of learning poverty gap and learning poverty severity are important to fully understand children¡¯s access to learning. It is possible that countries with the same learning poverty level have different learning poverty gaps, or countries with the same learning poverty gaps have different learning poverty severity, with implications for policies used to address learning poverty.

For example, where two countries have the same level of learning poverty, but one has a higher learning poverty gap, the latter would need greater effort to bring children above the minimum proficiency level. At the same time, where two countries have the same learning poverty gap, but one has higher learning poverty severity, the latter would need to adopt strategies that address the unequal distribution of learning among those below the minimum proficiency threshold. Furthermore, as we anticipate learning losses due to the pandemic, or the growing share of children who are learning poor, we can examine widening inequalities with the gap and severity calculations.

Calculation details

The implementation of this indicator and the production of the global estimates rely on:

• Reporting window of 9 years, a ¡À4 interval around a reference year. In the first release of the learning poverty, the reference year was set to 2015, implying data from 2011-2019 could be included. In practice, most recent data was from 2017.
• Learning assessments with a minimum proficiency threshold benchmarked by Global Alliance to Monitor Learning (GAML), which occurred within the reporting window. If a country has multiple eligible learning assessments, the following hierarchy is applied: PIRLS reading > TIMSS science > Regional assessments > National assessments. Between two rounds of the same assessments, the one closest to the reference year is preferred.
• School participation is derived from adjusted net enrollment rate (ANER) for primary schools and computed by the UIS using administrative records. Adjusted net enrollment is a measure of both ¡°stock¡± and ¡°flow¡± and accounts for both age- and grade-based distortions, as it is the percent of primary school age children enrolled either in primary or secondary education, as opposed to gross enrollment which is the share of children of any age that are enrolled in primary school, or net enrollment which is the share of primary school age children that are enrolled in primary school. We use the same year of school participation as the preferred learning assessment for each country.
• Aggregations for each region comprise the average learning poverty of countries with available data, weighted by their population ages 10¨C14 years old. To obtain a global estimate, we weight the regional aggregations by the 10¨C14-year-old population regardless of data availability. This is equivalent to imputing missing country data using regional values.

Note: While the reference age for Learning Poverty is age 10, learning assessments are sampled based on specific grades and not age. To incorporate assessments administered at different grades, we chose for each country the grade between 4 and 6 where relevant and reliable data were available.

Data

You can download the Learning Poverty data directly from . The database contains pooled and gender-disaggregated indicators for percent of children in learning poverty, percent of primary school-aged children who are out of school, and percent of children below minimum proficiency in reading at the end primary.

You may also access the learning poverty data directly through .

To load the Learning Poverty data directly in Stata you can use this code:

// Install the user-written command if you don't have itcapture which wbopendataif _rc == 111 ssc install wbopendata // Query Learning Poverty indicator from World Bank APIwbopendata, indicator(SE.LPV.PRIM) latest long clear

To load the Learning Poverty data directly in Python you can use this code:

# Load the packageimport wbgapi as wb # Query the most recent non-empty value (mrnev parameter)df = wb.data.DataFrame('SE.LPV.PRIM', db=12, mrnev=1, columns='time', numericTimeKeys=True)

Current findings

Learning poverty map

The map below is a snapshot of Learning Poverty across the world. You can also view the indicator for females and males. You may edit this map directly in .

How does learning poverty vary by gender?

Using all available cross-country assessments (as well as gender-disaggregated enrollment data from UIS), we have computed gender-specific learning poverty rates. Given data availability, we have only been able to compute this disaggregation for 92 countries. Access to microdata in some countries, particularly in South Asia, has been a significant challenge to compute gender-disaggregated outcomes.

^{Learning Poverty gender gap, by country}

Despite the barriers confronting girls in some areas of education, in virtually all countries for which we have data, girls have lower rates of learning poverty than boys do.

Replicate our results in GitHub

Our processes are documented on the  Github repository, which also includes instructions on how to  our numbers. You can find information about data source selection, calculations, aggregations .

Forthcoming update

The recent release of new learning assessment results ¨C TIMSS 2019, SEA-PLM 2019, and PASEC 2019 ¨C calls for an update of the learning poverty indicator. A public update of the regional and global estimates is planned for September 2021, to include the forthcoming LLECE 2019 results.

Significant changes are anticipated in some country estimates due to the replacement of national learning assessments by international ones. The initial learning poverty estimate was 52.7 percent in low- and middle-income countries, anchored in 2015. It used data from 62 countries, covering 80 percent of the target population. In September 2021, we plan to publish a corporate update of these global numbers. Using 2017 as the reference year implies accepting assessments from 2013 onwards, including the recently released TIMSS, SEA-PLM, PASEC from 2019 and the forthcoming LLECE 2019. With the new data, the coverage of the indicator will increase to 66 countries and 81 percent of the target population. The new update will also allow temporal comparisons in instances where countries have results from the same assessment in the last round.

Learning Poverty serves as an early-warning indicator for the Human Capital Project. For more information, visit website.