S pe Ed ci uc ResearchDigest al a Is tio su n World Bank e on VOLUME 12 NUMBER 1 FALL 2017 Realizing Education’s Promise A new report focuses on improving these severe shortfalls constitute a IN THIS ISSUE the quality of education as a means learning crisis. to equip young people for life and The report argues that low learn- Realizing Education’s Promise … page 1 ing is not inevitable; when improving work in a time of rapid change Investing in School Readiness … page 2 learning becomes a priority, great S progress is possible. In the early 1950s chooling is not the same as the Republic of Korea was a war-torn Can Participation by Parents Improve learning, emphasizes the World society held back by very low literacy School Quality? … page 3 Bank’s newly released World levels; by 1995 it had achieved univer- Development Report 2018. In Kenya, sal enrollment in high-quality second- Does More Time in School Improve Learning Outcomes Tanzania, and Uganda, when grade 3 ary education, with its young people for Students? … page 4 students were asked recently to read performing at the highest levels on a sentence such as “The name of the international learning assessments. How Does Partnering with the dog is Puppy” in English or Kiswahili, Vietnam surprised the world when the Government Affect Low-Cost Private three-quarters of the students did 2012 results from international assess- Schools? … page 5 not understand what it said. In rural ments showed that its 15-year-olds India nearly three-quarters of stu- performed at the same level as those Making Teachers Better: What We dents in grade 3 could not solve a in Germany—even though Vietnam Know and How to Learn More … page 6 two-digit subtraction such as 46 −17, was a lower-middle-income country. and half still could not do so by grade Between 2009 and 2015, as a result of Improving Service Delivery for Better 5. Although the skills of Brazilian concerted policy action, Peru achieved Education and Health Outcomes 15-year-olds have improved, at their some of the fastest growth in overall in Uganda … page 7 current rate of improvement they will learning outcomes. And in Liberia, not reach the OECD average score in Papua New Guinea, and Tonga early- Building More Schools: The Effects math for 75 years. In reading it will grade reading improved substantially on Equality of Opportunity in Education … page 8 take more than 260 years. And these within a very short time, thanks to fo- are all countries that have measured cused evidence-based efforts. Predicting School Dropouts in learning and made the results public; The learning crisis has three main Guatemala and Honduras … page 9 in too many other countries the prob- dimensions. First, the poor learning out- lem remains hidden. comes themselves: low levels of learn- Teachers’ Effectiveness in Pakistan Schooling without learning is not ing, high inequalities (across income, and the Link to Wages … page 10 just a wasted opportunity, but a great gender, and other characteristics), injustice: the children whom society and slow improvements in learning. Going Beyond Schooling to Improve is failing most are the ones in greatest Second, the immediate causes of the Workforce Skills … page 11 need of a good education to succeed crisis, seen in the various ways that in life. Within countries, learning out- the teaching-learning relationship Recent Policy Research Working comes are almost always much worse breaks down—such as when students Papers on Education … page 12 for the disadvantaged. In Uruguay poor are hobbled by a lack of early nutri- children in grade 6 are assessed as tion, teachers are unprepared or un- “not competent” in math at five times motivated, materials and technology the rate of wealthy children. Moreover, don’t improve learning, and school such data are for children and youth lucky enough to be in school. Together, (continued on page 12) 2 World Bank ResearchDigest Investing in School Readiness Children in rural Indonesia can is considerable variation in the path- centered on making a single year of follow different pathways of early ways of early childhood education. And preprimary education compulsory, childhood education. Which of the unsurprisingly, the authors find that these findings are timely. these pathways are highly correlated In the Indonesian context, the pathways is most cost-effective? with family and community character- added benefit from enrolling in play- M istics. For example, mother’s education group then kindergarten is probably any developing countries are level, family income, and the quality due to the different curricula used in increasing their investments of early childhood services in the com- each of these program settings. Child in early childhood education munity are all positively correlated development research has shown that because it improves children’s readi- with the probability of children’s en- children’s development is best sup- ness for school. Given the wide range rollment in both playgroup and kinder- ported when they receive increasingly of programs that already exist in some garten rather than only in playgroup or complex, differentiated learning ex- countries—and even in local settings only in kindergarten. periences. Play-based learning helps within countries—one of the chal- This self-selection into different children develop their fine and gross lenges facing policy makers is decid- pathways of early childhood educa- motor skills, develop language and so- ing what to fund. tion has potential longer-term con- cialization skills, and become creative Policy makers might begin by look- sequences. Using data collected as problem solvers. But the playgroups in ing at the existing evidence on effec- part of an impact evaluation in rural the study were not designed to provide tive early childhood programs. But Indonesia, the authors examine how multiple years of unique, developmen- the vast majority of research efforts the sequence, timing, and duration of tally appropriate learning. As a result, measuring the effectiveness of early attending different types of early child- children who subsequently enroll in childhood education in developing hood education (between ages three kindergarten are more likely to avoid countries are evaluation studies. As a and five) predict subsequent academic redundancy in their learning experi- result, they typically compare children outcomes in primary school (between ences by having exposure to different, who attended preschools with those ages six and nine). The primary school more academically focused curricula. who never attended—or compare test score of children who enrolled in The results of the authors’ analy- children who attended preschools playgroup then kindergarten is about sis show significant disparities in with improved quality to those who one grade level above that of their early learning outcomes by different attended unimproved preschools. peers who never enrolled in any early pathways. Children from the most Comparisons like these provide childhood education (with child, fam- socially disadvantaged backgrounds evidence supporting investment in ily, and community characteristics held were significantly less likely to receive high-quality preschools. But they may constant). Moreover, children who adequate exposure to a combination not adequately capture the reality of enrolled in the full sequence of both of play-based learning and academic many local settings where children can playgroup and kindergarten were about preparation to help them succeed in enroll in different types of early child- one semester ahead of their peers who primary school. Yet it is precisely these hood programs, at different ages, and enrolled only in kindergarten or only in children for whom this intervention for different lengths of time. playgroup. is likely to be most effective. The re- Take Indonesia, where a variety Enrollment in playgroup then sults suggest the importance of policy of early childhood services are avail- kindergarten also appears to have aimed at leveling the playing field in able. Some offer primarily play-based advantages for cost-effectiveness. For early childhood education—to ensure care (kelompok bermain, or playgroups), every $100 spent on early childhood that children from the most socially while others focus more on academic education, there are measurable gains disadvantaged backgrounds arrive at preparation for primary school (taman in primary school test scores only primary school ready to learn and have kanak-kanak, or kindergartens). Despite when children enroll in a combination an equal chance of performing well. formal guidelines about the intended of playgroup and kindergarten at the age group and length of programs, intended ages—that is, when the chil- families make personal decisions dren enroll in playgroup at ages three about early education pathways and and four followed by kindergarten at choose where, when, and for how long ages five and six. This suggests that Nozomi Nakajima, Amer Hasan, Haeil Jung, to enroll their children in early child- focusing on providing young children Sally Brinkman, Menno Pradhan, and Angela hood programs. with access to both playgroups and Kinnell. 2016. “Investing in School Readiness: An A recent paper by Nakajima, Hasan, kindergartens at the appropriate ages Analysis of the Cost-Effectiveness of Early Child- Jung, Brinkman, Pradhan, and Kinnell can optimize public investments in hood Education Pathways in Rural Indonesia.” shows that even among a sample of early childhood education. With the Policy Research Working Paper 7832, World poor, rural villages in Indonesia, there current policy discussions in Indonesia Bank, Washington, DC. World Bank Research Digest 3 Can Participation by Parents Improve School Quality? A study in Niger shows that grants responsibility: communities with the response to the grant. This may be be- can increase parent participation grant participated more and took on cause some teachers have a preference in school management—but this more responsibilities than those with- for a centralized government and were out the grant, though most did not reluctant to collaborate with parents, increased participation may not monitor teacher attendance. Parents especially when parents did not spend improve school quality did not reduce their own contributions the grant money on projects that made T in response to the grant. the teachers’ lives easier. hanks to an unprecedented effort The impact on school manage- The authors highlight barriers that to expand education, access to ment was mixed. Cooperation between communities may face when making schooling has increased dramati- school stakeholders improved, but school management decisions. Parents cally in poor countries over the past overall accountability did not change. will not always make optimal spending two decades. But the quality of educa- Spending showed both expected and and management decisions to improve tion is often low. One common strate- unexpected changes. There was more quality—in particular, parents may lack gy for improving quality is to increase spending on infrastructure, but also on information about how schools work. the participation of parents and the school festivals and playground equip- Cooperation (rather than confronta- community in school management ment. Most unexpectedly, there was tion) between parents and teachers and oversight. Community-based increased investment in agricultural may be key, but it can be difficult to management programs have been im- projects, which were probably not edu- establish. Finally, putting pressure on plemented in more than 20 countries, cational but intended to make a profit. teachers to improve service quality both developed and developing. However, different school contexts may be very difficult because formal The success of these programs de- led to different impacts. First, in situa- authority (the right to make decisions) pends on parents having the time, en- tions where the school committee was need not imply real authority (effective ergy, and capacity to effectively partici- educated or had experience in another control over decisions); parents’ au- pate in school management. Given the community organization—both of thority depends on their power relative cost of the programs, it is important to which are proxies for real authority— to teachers. understand under what circumstances parents increased their monitoring of Community participation may not those conditions might be met. teacher attendance in response to the be effective if communities have no Based on a field experiment in grant (though this did not mitigate actual power. Authority and capacity Niger, a recent study by Beasley and the negative effect of the grants on are important prerequisites for parents Huillery provides evidence that par- teacher attendance). Second, in small to successfully undertake the difficult ents were willing to increase their (one-teacher) schools, school commit- aspects of management and improve participation in school management, tees spent on items that benefited the school quality. but that this participation did not teacher, and in these schools teacher necessarily lead to improved educa- attendance increased in response to tional quality. The experiment tested the grant. These results together sug- a program in which the Ministry of gest that teachers’ response to par- Education, working in partnership with ents’ participation depends on whether the World Bank, gave grants to school parents are acting in opposition to committees with the aim of increasing the teachers or in alliance with them. parents’ involvement in school man- Third, while rural schools used some agement. The program took place in of the grant to invest in agricultural rural regions of the country, a context opportunities, urban schools did not where parents have limited capac- but invested in school infrastructure ity and authority because education instead. levels among adults are extremely low School quality did not improve and the education system is hierarchi- with these changes in the short term. cal and centralized. The experiment There were subsequent improvements involved 1,000 schools, of which 500 in infrastructure and health resources received the grant and the other 500 as well as an increase in participation did not (serving as controls). The first at the lowest grades, with fewer drop- grant arrived in late 2007, and the outs in 2007/08 and higher enrollment Elizabeth Ruth Beasley and Elise Huillery. 2017. evaluation ended in 2009. in grade 2 in 2008/09. But there is no “Willing but Unable? Short-Term Experimental The results show that, overall, evidence of a change in test scores. Evidence on Parent Empowerment and School the grant program had a positive Moreover, there is some evidence Quality.” Policy Research Working Paper 8125, impact on parents’ involvement and that teachers reduced their effort in World Bank, Washington, DC. 4 World Bank ResearchDigest Does More Time in School Improve Learning Outcomes for Students? Simply extending the time students The certified model also includes authors find in the Single-Shift schools spend at school may do little to an innovative curricular matrix as well appear to come from an artificial de- improve their learning outcomes. The as focused in-service teacher training crease in age-grade distortion (that is, to prepare teachers for this new ap- in the proportion of students who are answer? Full-time teachers proach. The curricular matrix has two more than two years older than the R parts. One consists of the common target age for their grade). io de Janeiro’s municipal network courses for all schools—including These results suggest that the certi- of public schools, the largest in Portuguese, math, and general cur- fied Single-Shift model implemented Brazil, combines different mod- ricular requirements. The other is a in Rio de Janeiro is successful in im- els of full-time education alongside diversified part—with directed study, proving student performance. But sim- the part-time schools that still make “life project” classes, and electives ply extending the daily time in school up the majority of school units. Taking designed in accordance with the ex- to seven hours does not seem to have advantage of this diversity of full- pertise of each school’s teachers. In the same positive impact on student time experiences, a recent paper by developing the certified model, Rio’s performance without a structured or- Cruz, Loureiro, and Sa examines the Municipal Secretariat of Education ganization for what is done in those impacts of these arrangements on stu- also sought to create schools that seven hours—such as that established dent learning outcomes. would develop youth leadership abili- through the certified model’s curricu- The full-time school program in ties through the curricular material, by lum, teacher training, and requirement Rio de Janeiro—called the Single-Shift using discussions and practices related that teachers dedicate 40 hours a week program—extends the time students to socioemotional skills. to the school. spend in school each day while also Applying a rigorous analytical ap- The paper adds to the literature seeking to improve the quality of the proach, the authors investigate wheth- on the impact of full-time schools in education provided. Unlike the model er students who have attended Single- developing countries. The findings of prevailing in most Brazilian public Shift schools had greater increases in previous studies do not indicate con- schools, in which the school day is learning than they would have had if clusive effects for the full-time school split into two shifts of about four hours they had not attended such schools. model; more time in school does not each (with each student attending one The authors use two indexes to mea- necessarily translate into student shift or the other), the Single-Shift pro- sure learning outcomes, the IDEB learning and performance. In con- gram uses a format in which students (the national Education Development trast, this paper finds that when the attend a seven-hour daily shift. Index) and the IDERio (a similar index extra time is organized with a struc- Beyond the seven-hour shift, some for the Rio de Janeiro municipality). tured curriculum, full-time teachers, of the Single-Shift schools have intro- Both indexes are based on student and focused teacher training, it may duced additional features and become performance on a standardized test have a significant impact on student certified (in 2015, of the 172 Single- (for the IDEB, the nationwide Prova achievement. Shift schools in Rio, 42 were certified). Brasil, taken in the fifth or ninth grade, One such feature is that all teaching and for the IDERio, the municipality- staff dedicate 40 hours of their weekly wide Prova Rio, taken in the third and working hours to a single school. seventh grades) as well as the passing (While having all teachers under 40- rate for the cycle (from first through hour contracts is a requirement for fifth grades or from sixth through certified middle schools, this is a goal ninth). rather than a requirement for certified The analysis indicates that the elementary schools.) certified Single-Shift program has had In certified Single-Shift schools, positive effects. The results are posi- teachers teach for seven hours daily tive and significant for certified Single- and collectively take one hour a day to Shift middle schools. For noncertified plan the next day’s lessons and discuss Single-Shift middle schools, however, issues relating to particular students the analysis shows nonsignificant or classes. That is, time is reserved results—and even negative results each day for all teachers to spend to- in some specifications. This suggests gether while planning their lessons. that merely extending the school day Tassia Cruz, Andre Loureiro, and Eduardo Sa. This feature of the certified model has does not generate positive impacts 2017. “Full-Time Teachers, Students, and Curric- resulted in changes in several other on student performance. The results ulum: The Single-Shift Model in Rio de Janeiro.” aspects, including teachers’ working for the elementary schools are less in- Policy Research Working Paper 8086, Impact hours and curricular guidelines. teresting. The positive effects that the Evaluation series, World Bank, Washington, DC. World Bank Research Digest 5 How Does Partnering with the Government Affect Low-Cost Private Schools? Governments are increasingly The paper shows that the PPP smaller; is more invested in children’s partnering with the private sector program absorbed large numbers of schooling (with parents reported to to provide education to low-income eligible students: student enrollment be more likely to visit their children’s in participating schools increased by school); and has more educated par- students. Is this a good policy? about 35 percent on average. Contrary ents (with students in PPP schools T to the hypothesis that PPP arrange- reporting a higher education level for he government of Uganda abol- ments might make private schools less their father). Students in PPP schools ished secondary school fees in responsive to parents, the paper finds also had better scores on the exams 2007. Over the next five years no adverse effects on the governance taken at the end of primary school. student enrollment in secondary of participating schools. While the pro- While the data are not well suited grades increased by 25 percent. But gram influences the topics discussed to reliably establishing who is driv- the number of classrooms and teach- at meetings (with head teacher and ing the “selection” of students, the ers in public schools increased at a teacher salaries discussed more often preponderance of evidence suggests much slower pace, leading to large in PPP schools), school ownership and that it is households and not schools. class sizes and high student-teacher control remain unchanged. And a PPP In other words, parents from more ratios. The situation is similar in arrangement increases the likelihood educationally favorable backgrounds several other Sub-Saharan African that a low-cost private school contin- are choosing to send their children to countries that have abolished school ues to operate (rather than shutting low-cost private schools participating fees for basic education. One com- down). in the PPP program. mon solution is to form public-private Moreover, partnering with the gov- What does this mean for policy? partnerships (PPPs) in which the gov- ernment appears to have a positive The paper shows that the PPP pro- ernment enters into contracts with impact on student performance in low- gram in Uganda is able to use the private schools to deliver services to cost private schools. Students enrolled private sector to help absorb a larger low-income households. for more than a year in a PPP school secondary student population with- Are PPPs a good solution? What have test scores in math, English, and out adversely affecting the quality of quality of education do these schools biology that are in the range of 0.07 to education. But the PPP program also provide to low-income students? And 0.16 standard deviations better than leads to a selection of students in how does partnering with the gov- those of students in nonparticipating participating schools who tend to be ernment change the private schools private schools. from more educationally favorable themselves? A recent paper by Barrera- Why do students perform better in backgrounds—a selection most likely Osorio, de Galbert, Habyarimana, and low-cost private schools that partner driven by households. This means that Sabarwal attempts to answer these with the government? The paper offers to understand the full implications questions by evaluating a PPP pro- two explanations. of the PPP program, further study is gram for secondary schools initiated in First, participation in the PPP pro- needed to investigate how it affects 2007 by the government of Uganda. gram leads to somewhat better teacher neighboring schools—both public and Uganda’s PPP program is open to performance and inputs in low-cost private. all certified private schools charging private schools. During unannounced low fees and meeting a set of quality visits a larger share of teachers in par- benchmarks. These low-cost private ticipating schools were found to be schools can enter into a contractual present in school and in class teach- arrangement with the Ministry of ing. In an environment where teachers Education under which they receive typically work in several schools at a a fixed grant per eligible student en- time, this means that being in a PPP rolled. To evaluate the impact of the program allows a low-cost private program on participating schools, the school to induce a greater teacher paper uses a randomized phase-in ap- commitment. The program also in- proach. A set of 101 private schools creases the likelihood of a school hav- that applied for the program in 2011 ing a science laboratory. were randomly divided into two Second, and more importantly, Felipe Barrera-Osorio, Pierre de Galbert, James groups: half received the PPP program once a low-cost private school enters Habyarimana, and Shwetlena Sabarwal. 2016. in 2011 (the treatment group), and the into a PPP arrangement, it appears to “Impact of Public-Private Partnerships on Private other half received the program only in attract students from more education- School Performance: Evidence from a Random- 2012 (the control group). The impact ally favorable backgrounds. Students ized Controlled Trial in Uganda.” Policy Research of the program was estimated by com- in PPP schools appear to be more Working Paper 7905, World Bank, Washington, paring outcomes in these two groups. likely to come from a household that is DC. 6 World Bank ResearchDigest Making Teachers Better: What We Know and How to Learn More Teacher training programs that for teachers in low- and middle- professional development for teachers matter for promotion and provide income countries. Yet in the course of would be much more effective. complementary materials are more that analysis the authors realized that Based on the 26 programs, what many evaluations do not report the do the most successful ones look like? effective. A new instrument will help characteristics of their programs in a Those that lead to greater student us continue to learn what works consistent way. The authors therefore learning gains include programs that T propose a standardized instrument have implications for teachers’ sal- he quality of education in low- for reporting on professional devel- ary or promotion. They also include and middle-income countries is opment programs for teachers—the programs that provide complementary lower than it should be. Teachers In-Service Teacher Training Survey materials to enable teachers to imple- are the most important input to stu- Instrument (ITTSI)—and use it with ment the methods in which they have dent learning. These two statements the 26 programs they identified. been trained. Other results (less defin- are well established. The question What characteristics are likely to itive because of the limited sample of emerges, How can the current set of matter? The instrument includes sec- programs) suggest that programs with teachers become better? tions on content (Does the training fo- follow-up visits to teachers in schools Programs providing on-the-job cus on upgrading math knowledge? Or lead to larger gains, while those with- training to teachers—known as profes- on pedagogical skills?), delivery (Is it a out a specific subject focus (“student- sional development programs—are lecture series or a mentoring program? centered pedagogy” rather than “peda- common interventions aimed at im- How long does it last?), overarching gogy for geometry,” for example) lead proving the quality of teaching. Recent aspects (Does it have any bearing on a to smaller ones. reviews of the evidence indicate that teacher’s professional status? What is Implementers also commonly re- some of the most promising interven- targeted at particularly low-performing port conducting follow-up visits and tions to improve student learning are teachers?), and even perceptions (What designing programs in response to such programs. But experienced pro- did teachers and trainers believe was local needs as key to effective profes- fessionals will report that much of the most and least effective about the sional development. All this comple- teacher training now being provided program?). Altogether, the instrument ments evidence from high-income is ad hoc, overly theoretical, and ulti- includes 73 simple indicators. countries, which indicates that more mately ineffective. Both are true. Some Evaluations varied dramatically in specific, practical training, sustained professional development programs what they reported (figure 1), so the over time and with follow-up visits have led to major learning gains, such authors interviewed as many program to teachers in schools, translates as one in Liberia that trained teachers implementers as possible to fill in the into larger improvements in student and gave them materials to work with, data. After interviews, information on learning. and one in Uganda that provided prac- virtually all the indicators was avail- Professional development for tical training and follow-up mentoring. able. In other words, implementers teachers uses an enormous amount But many probably deliver no gains at know all these characteristics of their of financing. Of more than 170 World all. programs; researchers just do not Bank projects with education com- In a recent paper Popova, Evans, report them consistently. If research- ponents in the early 2000s, nearly and Arancibia sought to identify the ers were to begin using the proposed two-thirds included professional de- common characteristics of successful instrument, future analysis of the velopment. Virtually every ministry of professional development programs latest evidence on what works in education dedicates resources to it. For governments, this suggests the im- Figure 1. Share of ITTSI Indicators Reported by the 26 Program Evaluations portance of ensuring that professional development translates into gains for students in the classroom. It also sug- gests the value of documenting their innovations so that future administra- tions and neighboring countries can Programs Average learn from their experience. Anna Popova, David K. Evans, and Violeta Aran- cibia. 2016. “Training Teachers on the Job: What Works and How to Measure It.” Policy Research 0 10 20 30 40 50 60 70 80 90 Working Paper 7834, World Bank, Washington, Percent DC. World Bank Research Digest 7 Improving Service Delivery for Better Education and Health Outcomes in Uganda Efforts to improve education and applied a rigorous methodology to to diagnose five key conditions in pa- health services in Uganda appear collect robust evidence on the quality tients and the accuracy of diagnoses to have paid off—though gains of primary education and basic health again weaker in poorer areas. services. To provide a basic diagnostic Another important finding is that continue to lag in poorer areas of the quality of these services, a new while the quality of service delivery as S paper by Tsimpo, Etang, and Wodon well as education and health outcomes ince 1997 the government of relies on the SDI survey as well as tend to be lower in poorer communi- Uganda has implemented a another nationally representative sur- ties, households in these communities series of policies to improve vey (the Uganda National Household tend to be at least as satisfied with the education and health outcomes, with Survey of 2012/13, which has a commu- services they receive as are households a focus on disadvantaged groups. On nity module providing data on educa- in better-off areas. This suggests that the supply side, policies have aimed tion and health facilities). expectations of quality and the ability to improve both the access to and The authors’ analysis suggests that to discern poor quality are likely to be the quality of services. These poli- despite the government’s efforts, the low in poorer areas, implying that sim- cies include building and renovating quality of basic services remains low, ply looking at satisfaction with services schools and health centers; purchas- especially for those living in poorer in traditional household surveys is ing adequate instructional materials; areas. This has implications for devel- unlikely to lead to robust measures of training, hiring, and retaining teachers opment outcomes. service quality. and health workers; ensuring the avail- While access to education has in- None of these findings are espe- ability of drugs in health centers; tack- creased nationally, learning outcomes cially surprising, and research for other ling absenteeism among teachers and remain weak, especially in disadvan- countries using SDI surveys tends to health workers; providing incentives taged areas. Students in poorer areas yield similar findings. But at the coun- for staff to work in areas hard to reach tend to perform less well on standard- try level, this type of analysis is useful or hard to stay in; and implementing ized tests for English, numeracy, and to identify aspects requiring improve- curriculum and other changes. nonverbal reasoning. This appears to ment in the provision of education and There have also been notable inter- be in part because of lower-quality health services, especially in disadvan- ventions on the demand side. These inputs in schools in those areas, in- taged areas. Furthermore, by repeating have included making primary and, cluding teaching practices that engage this type of survey over time, countries later, secondary schooling free; imple- students too little and a limited knowl- can assess their progress in improving menting and expanding school feed- edge base among teachers. Teacher services in more robust ways. The good ing programs; launching the “mama absenteeism—when a teacher either is news is that the analysis of the data kit” program to reduce out-of-pocket not in school or is in school but not in for Uganda suggests that better quality costs for deliveries in health facilities; the classroom—is also a concern. of service delivery does indeed tend and rolling out national immunization Similar issues are observed in the to lead to better education and health days. health sector, though with some dif- outcomes for the population. While there has been progress in ferences. The correlation between the many areas, challenges remain. In welfare level of communities and the particular, there is a concern that the quality of some of the inputs in health increase in access may have been facilities is weaker than for education. achieved at a cost in terms of the qual- As an example, unlike what is ob- ity of the services being provided. As served in schools, there is no apparent a result, quality of service delivery has correlation between health workers’ become a concern for many stakehold- absenteeism and the welfare level of ers, including households and govern- communities, and the availability of ment and development partners. drugs is not necessarily lower in poorer Until recently, limited evidence areas. On the other hand, substantial was available to support the policy differences remain in the availability debate on quality of service delivery in of electricity, piped water, toilets, an Uganda, especially in education and ambulance, a microscope, a thermom- health. To fill the data gap, the World eter, a weighing scale, a blood pressure Clarence Tsimpo, Alvin Etang, and Quentin Bank together with partners imple- machine, or kits to test for malaria Wodon. 2017. “Education and Health Services mented a Service Delivery Indicator or HIV—with lower-quality inputs in in Uganda: Quality of Inputs, User Satisfaction, (SDI) survey in 2013. As in other coun- poorer areas. And as is the case for and Community Welfare Levels.” Policy Research tries where the SDI survey has been education, health workers’ knowledge Working Paper 8116, World Bank, Washington, implemented, the survey in Uganda is limited, with only a small share able DC. 8 World Bank ResearchDigest Building More Schools: The Effects on Equality of Opportunity in Education A new study finds that an increase time-invariant characteristics of sub- go to school in a jurisdiction different in local schools in Jordan has led to districts, period effects that differen- from their own. greater intergenerational mobility in tially affect individuals in different co- With respect to local secondary horts, and the time-varying character- schools, the authors find that a greater education—especially for women istics of the governorate of birth that supply increases schooling attain- J may drive both the local public school ment for male children of uneducated ordan has pursued policies supply and the educational attainment parents, with the effect stemming aimed at greatly expanding its of individuals in a given year of birth. from both boys’ and mixed second- supply of public basic and sec- By first analyzing intergenerational ary schools. But it has no statistically ondary schools, and the country had mobility across cohorts, the authors significant effect for females. Moreover, one of the world’s biggest increases in establish that mobility has increased the effect shows no variation across educational attainment between 1980 significantly in Jordan over time, and different levels of parents’ education and 2010. But has the increase in the more so for women than for men. And for either males or females. This sug- local supply of schooling improved they find that the increased supply gests that in the absence of public the equality of opportunity in edu- of public basic schools has improved secondary schools, even educated par- cation? That is, has it led to greater intergenerational mobility, especially ents have few options for sending their educational attainment even among among females, but that the increased children to secondary school, given those whose parents are relatively supply of secondary schools has not the relative absence of private second- uneducated? done so. ary schools outside the capital city A recent paper by Assaad and Saleh Indeed, the authors find that an of Amman. Thus both educated and examines how the enhanced local sup- increase in the local supply of basic uneducated parents are equally con- ply of schooling in Jordan has affected schools reduces the intergenerational strained by the local supply of public the association between children’s persistence of educational attainment secondary schools. schooling and that of their parents—or for women three times as much as it This research shows that a more the intergenerational mobility in edu- does for men. Increasing the supply of progressive government policy to con- cation. The authors use a unique data sex-appropriate basic schools by one struct more public basic schools and source, the 2010 Jordan Labor Market standard deviation (approximately 0.19 to equalize the supply of basic schools Panel Survey, which includes informa- schools per 1,000 people for males across jurisdictions does in fact con- tion on parents’ schooling for every and 0.13 schools per 1,000 people for tribute to greater equality of opportu- adult in the sample, along with the females) reduces the father-son and nity in education. It remains for future 2010 school census produced by the mother-son associations of schooling research to determine whether it is Ministry of Education. The school cen- attainment by 18–20 percent, while it better in a socially conservative setting sus provides the subdistrict, type, and reduces the father-daughter associa- such as Jordan to establish single-sex date of establishment for every school tion by 44 percent and the mother- schools or mixed schools if the objec- in Jordan, making it possible to mea- daughter association by 33 percent. tive is to improve the educational at- sure the local supply of each type of Moreover, the increased supply of tainment of girls. school in each subdistrict in every year. basic schools significantly enhances Individuals’ exposure to the supply of the schooling attainment of the chil- public schooling is then determined dren of uneducated parents, with the by the number of sex-appropriate ba- effect being larger for females than for sic (or secondary) public schools (per males. These effects are due primarily 1,000 individuals) available to them to the availability of mixed schools for in their subdistrict of birth at the time both boys and girls. they were of the appropriate age to en- Overall, the findings suggest that roll in that school level (age 6 for basic the local supply of basic schools is and 15 for secondary). The richness of a binding constraint on both the the data set makes it the first in the educational attainment and the inter- Middle East to allow such a study. generational educational mobility of The authors identify the effect of lo- Jordanians, with the constraint being cal school supply on intergenerational more binding among females than mobility in education by exploiting males. The larger effect for women can Ragui Assaad and Mohamed Saleh. 2016. “Does the variation in the supply of public be explained by the fact that girls in Improved Local Supply of Schooling Enhance basic and secondary schools across a conservative social setting such as Intergenerational Mobility in Education? Evidence cohorts and subdistricts of birth. Their that of Jordan are more constrained from Jordan.” Policy Research Working Paper analysis controls for all unobserved geographically and are often unable to 7825, World Bank, Washington, DC. World Bank Research Digest 9 Predicting School Dropouts in Guatemala and Honduras Using administrative data from a out. This research is based largely on based on routinely collected data and developing country, a fairly simple the rich administrative data available relatively simple analytical techniques, model can predict which students in the United States, where more than could feasibly be implemented in a 30 states have some form of early wide range of country contexts. are most likely to drop out warning system in place. Similarly, By providing an accurate means of I most European countries monitor targeting, these models could substan- n countries across Latin America, early warning signs of potential drop- tially improve the cost-effectiveness primary enrollment has become out through their information manage- of dropout prevention programs. In a almost universal and secondary ment systems, primarily at the school simple simulation of a modest pro- enrollment has grown substantially level but in some cases nationally. gram, the authors show that using the in the past two decades. But attain- Solving the problem of how to models to target students, rather than ment continues to fall short of aspira- predict dropouts is now also becom- targeting poor municipalities or high- tions. The high rates of enrollment ing possible in many middle-income dropout schools, could reduce the in early grades quickly decline as countries that have invested in setting misallocation of resources by between students drop out before completing up information management systems 30 and 80 percent. a full course of basic education. In in recent years—through the use of Previous studies have provided Guatemala and Honduras, for exam- routinely collected student-level data. substantial evidence that dropout can ple, education is compulsory through In both Guatemala and Honduras stu- be accurately predicted using high- the ninth grade, but nearly 40 percent dent- and school-level data are now quality administrative data in the of sixth graders drop out before get- digitized in networked administrative United States as well as Europe. But ting there. databases thanks to the efforts of the no previous studies are known to have The high dropout rates raise impor- ministries of education in these coun- investigated this question in a devel- tant concerns. Global evidence sug- tries. These data allow the tracking of oping country. The authors’ paper thus gests that, on average, dropouts earn students over time and, in Guatemala, provides one of the first applications less and experience more social and can be directly linked to standardized of dropout prediction methodologies economic challenges than their peers test data. to the education systems of lower- who have completed more years of A new paper by Adelman, middle-income countries. education. Haimovich, Ham, and Vazquez uses Policy makers’ ability to effectively these administrative data from address these consequences depends Guatemala and Honduras to estimate on their ability to answer a fundamen- early warning models of dropout for tal question: Who is most likely to primary and secondary school stu- drop out? This question may appear dents. This involves two steps. First, easy to answer in countries with high the authors estimate linear probabil- dropout rates, because one might as- ity models to predict the likelihood sume that dropouts are concentrated of dropout for each student. Second, in particularly disadvantaged or dys- they employ an algorithm to select the functional schools or among students probability cutoff used to identify stu- with particular characteristics. But dents most at risk of dropping out. The dropouts are often spread across optimal cutoff is chosen to minimize schools and not readily identifiable by the distance to the “perfect prediction” single characteristics. In Guatemala (with 100 percent true positives and no more than half the sixth-grade stu- false positives). The authors also con- dents who drop out in the transition to duct different validation exercises that lower secondary are spread across the show that their models perform quite 70 percent of primary schools in which well out of sample. the dropout rate is below 50 percent. Using this methodology on stu- And while half the students who score dent-level panel data, the authors are in the lowest quartile on a sixth-grade able to accurately identify about 80 standardized exam drop out, so do percent of the sixth-grade students Melissa Adelman, Francisco Haimovich, Andres 20 percent of those who score in the who dropped out within the next year Ham, and Emmanuel Vazquez. 2017. “Predicting highest quartile. in Guatemala and Honduras—a level School Dropout with Administrative Data: New But recent research is providing an of performance comparable to those Evidence from Guatemala and Honduras.” Policy increasingly sound empirical base for of models used in the United States. Research Working Paper 8142, World Bank, accurately predicting who will drop These early warning models, which are Washington, DC. 10World Bank ResearchDigest Teachers’ Effectiveness in Pakistan and the Link to Wages Teachers are key to the quality of often used in teacher recruitment—is performed better than permanent learning, but their effectiveness is associated with value added. ones. difficult to predict and not linked to Are there teacher characteris- The fact that reducing teacher tics that predict higher salaries in salaries by 35 percent had no adverse their wages in the public sector Pakistan? The authors examine this effects on student outcomes might E issue in both the public and private seem surprising. But comparing public xtensive research from the United sectors. Results show that having a and private sector teachers’ salaries States clearly establishes the bachelor’s degree positively predicts helps illustrate why this is so: there is importance of good teachers for wages in both sectors. And in the pri- virtually no overlap, with public sec- students’ learning outcomes. It also vate sector, teachers with higher value tor teachers making about five times shows that predicting teacher qual- added are paid more. In the public as much on average as private sec- ity is difficult. Unfortunately, we don’t sector, however, there is no relation- tor teachers. This large public sector know whether these findings also ap- ship between teacher value added and premium is not unique to Pakistan. ply to low-income countries. Filling wages. These findings suggest that According to a recent study using data this gap is critical: student learning in teacher value added is at least some- from 52 countries, these premiums are these countries is poor, and teacher what observable and that the private larger in low-income countries than in salaries account for 80 percent of sector identifies and rewards better high-income ones. recurring education expenditures. teachers with higher wages. In con- Altogether, these results suggest Increasingly, countries are asking trast, higher wages in the public sector that calls to improve student learning how to recruit and retain high-quality are not associated with higher teacher by raising teacher wages, while keep- teachers. value added. ing the same hiring and pay structure In a new paper Bau and Das in- This lack of an association between in place, are misguided. Redesigning vestigate these issues using a unique teacher productivity and wages sug- compensation systems for government data set on primary school students gests that the public sector does not employees is complex enough as it is; and their teachers in both public and reward better teachers with higher one issue that policy makers need not private schools in Punjab, Pakistan, wages. But this does not mean that worry about is that teacher salaries are for 2004–07. The data track students increasing teacher wages would not too low to attract talented teachers. through primary school, matching attract higher-quality applicants for them to the teachers who taught them teaching positions. in each year. One novel aspect of the A change in hiring practices in data set is that it includes scores on Punjab makes it possible to assess primary-school-level tests in Urdu, how teacher wages affect the quality of English, and math for both students applicants. In 1998 Pakistan initiated and teachers. unanticipated nuclear tests, which Using student test score data, the led to international sanctions on the authors calculate a value added (as a country and tighter budget constraints proxy for productivity) for each teacher: in the province. As a result, teacher an estimate of the test score gains that hiring was largely frozen for three a random student would receive if as- years. When hiring resumed, almost signed to that teacher. These estimates all new teachers were hired under tem- show that moving a student from a porary contracts with 35 percent lower teacher at the 5th percentile to one at salaries. With the caveat that the new the 95th would increase the student’s teachers’ contract status may have test scores by 0.5 standard deviations increased accountability, the authors —the equivalent of more than a year compare the productivity of teachers of school. hired right before the nuclear tests So teachers clearly matter. But as with that of teachers hired right after Natalie Bau and Jishnu Das. 2017. “The Misal- in the United States, predicting their hiring resumed to see whether lower location of Pay and Productivity in the Public effectiveness is hard. The authors find salaries attracted worse teachers. Sector: Evidence from the Labor Market for Teach- that only teachers’ content knowledge The results show no negative ef- ers.” Policy Research Working Paper 8050, World (as measured by their scores on the fect of being hired at a lower salary Bank, Washington, DC. primary school tests) and their first on teacher productivity. Indeed, when This article is adapted from a longer version that two years of experience predict their the authors account for the fact that appeared in VoxDev on September 8, 2017, at value added. Neither a bachelor’s de- teachers are observed at different lev- http://voxdev.org/topic/health-education/labour gree nor teacher training—two factors els of experience, contract teachers -market-teachers-pakistan-pay-and-effectiveness. World Bank Research Digest11 Going Beyond Schooling to Improve Workforce Skills What does it take to succeed in complexity of computer use on the job Estimating a model that includes the labor market in developing as proxies. Literacy proficiency is de- all types of skills, the study finds that countries? Cognitive skills matter fined by a range of skills, from the de- computer skills continue to matter coding of written content to the com- most for labor market success, though above and beyond educational prehension, interpretation, and evalu- there is substantial variation across attainment ation of texts of varying complexity. For countries. In contrast, the association D the complexity of computer use, four between literacy proficiency and earn- riven by evidence that a skilled skill levels are defined, ranging from ings remains positive and significant labor force is critical to eco- browser-based tasks to basic Microsoft only for Ukraine. The association be- nomic development, low- and Office functions, basic programming, tween noncognitive skills and earn- middle-income countries worldwide and advanced programming. To mea- ings is slightly smaller relative to the are investing in strategies to improve sure noncognitive skills, the survey in- results of previous models used in the workforce skills. Beyond the benefits cludes a series of questions to assess study, but cross-country patterns re- at the national level, are there eco- personality and behavioral traits—the main qualitatively similar. nomic payoffs for individuals from “Big Five” personality traits (openness, The study’s findings show that cog- being more skilled? And are certain conscientiousness, extraversion, agree- nitive skills matter above and beyond types of skills associated with higher ableness, and neuroticism) as well as educational attainment—skills en- returns? grit and decision making. compassing additional dimensions of To assess the economic returns As proxied by literacy proficiency human capital not usually captured by to skills, earlier econometric studies and the complexity of computer use, traditional measures of schooling. The commonly used individuals’ level of overall levels of cognitive skills are findings underscore the importance of schooling as a proxy for skills. But us- low in the eight countries. Among education systems that can produce ing schooling measures alone system- the sample population, average lit- graduates who are proficient in com- atically underestimates cognitive skills eracy proficiency is more than one prehending, interpreting, and analyz- learned outside school and misses the standard deviation below the OECD ing written texts—skills that provide a contributions of other types of skills. average, and about half of those who foundation for lifelong learning as well More recent studies, based mainly in use a computer at work utilize only as being valued in the labor market. OECD countries, go beyond schooling basic programming functions such And they point to the need to ensure to measure cognitive or noncogni- as spreadsheets or bookkeeping. For that the future workforce develops the tive skills. But this newer evidence on noncognitive skills, average scores are digital skills to thrive in the workplace returns to skills, especially those nec- relatively stable across countries as as digital technologies continue to essary for success in technology-rich well as among different demographic spread and countries move toward the environments, is scarce in low- and groups within each country. technology frontier. middle-income countries. Analysis shows that skills yield The study also reveals important A recent paper by Valerio, Sanchez significant payoffs in the labor market, differences across countries in how Puerta, Tognatta, and Monroy-Taborda even after controlling for education different types of skills are valued in adds to the evidence in developing and other relevant individual and fam- the labor market. This suggests a need countries and contributes to an emerg- ily background factors. On average, for further research to provide a more ing literature investigating the dif- an increase in literacy proficiency of nuanced understanding of the role of ferent contributions of cognitive and one standard deviation is associated skills in labor market success and to noncognitive skills. Using data from with an increase in hourly earnings support strategies for developing work- the Skills towards Employability and ranging from 9 percentage points (in force skills that take into account each Productivity (STEP) survey, the study Colombia, Georgia, and Ukraine) to 19 country’s conditions, needs, and goals. investigates the extent to which there (in Ghana). Similarly, more complex is an independent association be- computer skills bring big earnings pre- tween earnings and skill types. It draws miums in all eight countries. In con- on survey data for eight countries: trast, the results for noncognitive skills Armenia, Bolivia, Colombia, Georgia, are less consistent across countries. Ghana, Kenya, Ukraine, and Vietnam. For example, openness, agreeableness, Its analytical sample includes all em- and grit show a significant and positive Alexandria Valerio, Maria Laura Sanchez Puerta, ployed and self-employed adults who relationship with earnings in five coun- Namrata Tognatta, and Sebastian Monroy- are between the ages of 25 and 64 and tries, while in another set of countries Taborda. 2016. “Are There Skills Payoffs in Low- are not in school. agreeableness and conscientiousness and Middle-Income Countries? Empirical Evidence To measure cognitive skills, the show an insignificant or negative asso- Using STEP Data.” Policy Research Working STEP survey uses literacy and the ciation with earnings. Paper 7879, World Bank, Washington, DC. 12World Bank ResearchDigest (continued from page 1) Recent Policy Research management is poor. Third, the deeper the NGO-led ASER and Uwezo pro- Working Papers on Education system-level barriers, both technical and grams have done in India and East political, that pull the various actors Africa) 7833 Gender Gaps in Cognitive and Non-Cognitive away from a focus on learning. The • Build coalitions to shift politi- Skills in Early Primary Grades: Evidence from Rural Indonesia report provides detailed diagnoses of cal incentives toward learning for all (as Nozomi Nakajima, Haeil Jung, Menno each dimension based on new data Chile did early in its decades-long Pradhan, Amer Hasan, Angela Kinnell, and Sally Brinkman and research. education reforms, or as Malaysia and 7843 Intergenerational Education Mobility in To do better, the report argues that Tanzania did recently with collabora- Africa: Has Progress Been Inclusive? Theophile T. Azomahou and Eleni A. Yitbarek a nation must take action on three tive society-wide reform “labs”) 7878 Do Cognitive and Noncognitive Skills Explain fronts. • Use innovative and adaptive ap- the Gender Wage Gap in Middle-Income Countries? An Analysis Using STEP Data First, assess learning, to make it a serious proaches to find out which approaches Namrata Tognatta, Alexandria Valerio, and goal. Countries need to put in place a work best in their context (as Burundi Maria Laura Sanchez Puerta 7881 Children Left Behind in China: The Role of range of well-designed student assess- did during postconflict reconstruction). School Fees ments to help teachers guide students, The payoff is education that de- Hai-Anh Dang, Yang Huang, and Harris Selod 7954 Can Solar Lanterns Improve Youth Academic improve system management, and livers: For individuals, it promotes Performance? Experimental Evidence from focus society’s attention on learning. employment, earnings, health, and Bangladesh Yuya Kudo, Abu S. Shonchoy, and Kazushi These measures can spotlight hidden poverty reduction. And for societies, Takahashi exclusions, inform policy choices, and it drives long-term economic growth, 7956 What Do Teachers Know and Do? Does It track progress. spurs innovation, strengthens institu- Matter? Evidence from Primary Schools in Africa Second, act on evidence to make schools tions, and fosters social cohesion. But Tessa Bold, Deon Filmer, Gayle Martin, Ezequiel Molina, Christophe Rockmore, Brian work for all learners. Countries should these benefits depend largely on learn- Stacy, Jakob Svensson, and Waly Wane start by targeting areas with the largest ing. Mounting evidence shows that 8014 Can Parental Migration Reduce Petty gaps between what happens in prac- the skills acquired are what equips Corruption in Education? Lisa Sofie Höckel, Manuel Santos Silva, and tice and what evidence suggests works individuals for work and life, and that Tobias Stöhr for learning. The best place to start is it is through learning and skills that 8071 Do Different Types of Assets Have Differential Effects on Child Education? these three key areas: prepared learn- education boosts growth. Countries Evidence from Tanzania ers; skilled and motivated teachers; have already made a start by getting so Kashi Kafle, Dean Jolliffe, and Alex Winter- Nelson inputs and management focused on many young people into school; now it 8112 Well Begun, but Aiming Higher: A Review teaching and learning. is time to realize education’s promise of Vietnam’s Education Trends in the Past 20 Years and Emerging Challenges Third, align actors, to make the whole by accelerating learning with equity. Hai-Anh H. Dang and Paul W. Glewwe system work for learning. Even evidence- 8135 Global Inequality in a More Educated World Amer Ahmed, Maurizio Bussolo, and Marcio based classroom innovation may have Cruz little impact if system-level technical 8156 Through the Looking Glass: Can Classroom and political factors prevent a focus Observation and Coaching Improve Teacher Performance in Brazil? on learning. Countries can escape low- Barbara Bruns, Leandro Costa, and Nina Cunha learning traps by acting on three fronts World Bank. 2018. World Development as they implement reforms: Report 2018: LEARNING to Realize Educa- Working Papers can be downloaded at http://econ.worldbank.org • Deploy information and metrics to tion’s Promise. Washington, DC: World Bank. To download the World Bank Research E-Newsletter, make learning politically salient (as http://www.worldbank.org/wdr2018. go to http://econ.worldbank.org/research_newsletter The World Bank Research Digest is a quarterly publica- The Research Digest is financed by the Bank’s Editorial Committee: Chorching Goh, Aslı Demirgüç- tion disseminating findings of World Bank research. Research Committee and managed by DECDP, the Kunt, and Shiva S. Makki. Editor: Alison Strong; The views and interpretations in the articles are those research support unit of the Development Economics production: Roula Yazigi. For information or free of the authors and do not necessarily represent the Senior Vice Presidency (DEC). The Research Digest is subscriptions, send email to research@worldbank.org views of the World Bank, its Executive Directors, or the not copyrighted and may be reproduced with appropri- or visit http://econ.worldbank.org/research_digest. countries they represent. ate source attribution. The World Bank 1818 H Street, NW Washington, DC 20433, USA Printed on Recycled Paper