what can we learn from education impact...
TRANSCRIPT
What can we learn from education impact evaluations?
Barbara Bruns Center for Global Development
SIEF Workshop
Dodoma, Tanzania October 4, 2017
Overview
1. Why evaluate? 3 reasons
2. What are we learning in education?
3. A birds’ eye view of our evidence base (from
Ganimian and Murnane) 3. Top ten policies based on evidence 4. Conclusions
Annex: Complete review of Ganimian/Murnane meta-study conclusions
To influence spending decisions: Mozambique’s Escolinhas
7,38
3,15
26,21
17,82
5,37 5,55
2,86
14,01
11,48
3,60
Physical Health Communication CognitiveDevelopment
Social Competence Emotional Maturity
Preschool Control
• Low-cost pre-school program developed by Save the Children (30 communities); 30% of cost of traditional pre-schools
• Volunteers from the community trained to work
with children aged 3-5, given small stipend, learning materials and weekly feedback
An impact evaluation showed significant effects on children’s development → Govt. decided to scale up to
600 rural communities
To influence program design
• Forces us to articulate our “theory of change” • Forces us to think through implementation steps –
Who will act? What resources will they need? How long will each step really take?
• Stimulates the collection of key monitoring data –eg. classroom observation
India (indiv.)
0.3-0.45
Kenya
Teacher bonus pay, ca. 2009
Bonus Pay Programs by Impact Size (learning outcomes)
0 0,1 0,2 0,3 0,4 0,5
Effect Size
Mexico ALI
(indiv.)
India (group)
5 yr. impact Israel (group)
Israel (indiv.)
India (attend.)
0.18
Chile (group)
0.14-0.26
Source: Making Schools Work, 2011
- Program design matters a lot
- Some programs with strong learning results did not have clear explanations of “how”
Theory of action: Results caused by a change in teacher policy must change teacher effort and/or effectiveness in the classroom
ACTIVITIES
Curriculum design
Teacher
training
School
supervision
INPUTS
Financing
Salaries
School
construction
Books
ICT/Materials
OUTPUT
Teacher
practice in
the
classroom
Inputs
Activities
Results
OUTCOMES
Student
learning
Graduation
Rates
Employment
Wage gains
Use standardized classroom observation to measure
Pernambuco teacher bonus: 2009-2011
ACTIVITIES
School targets
for improvement
in test scores
and pass rates
School bonus for
reaching targets (one month salary
for all teachers in
school)
INPUTS
Financing (10%
increase in salary
bill)
OUTPUT
Teachers
absent less
and spend
more time on
instruction
Inputs
Activities
Results
OUTCOMES
Pass rates
improved more
than learning
Lowest-
performing
students had
biggest gains
Stallings method classroom observations in 20% of
schools
More time on instruction correlated with bonus pay
awards
** Statistically significant at 5%
*** at 1%
Pass rates drove bonus awards; learning gains modest, but largest for low-performing students
Pernambuco IDEB gains (secondary education)
Santa Catarina 3.5 Pernambuco 3.9
Minas Gerais 3.4 São Paulo 3.9
Rio Grande do Sul 3.4 Goiás 3.8
São Paulo 3.3 Espírito Santo 3.7
Paraná 3.3 Paraná 3.6
Roraima 3.2 Rio de Janeiro 3.6
Espírito Santo 3.1 Acre 3.5
Acre 3 Amazonas 3.5
Distrito Federal 3 Distrito Federal 3.5
Ceará 3 Mato Grosso do Sul 3.5
Rondônia 3 Minas Gerais 3.5
Goiás 2.9 Ceará 3.4
Tocantins 2.9 Roraima 3.4
Rio de Janeiro 2.8 Santa Catarina 3.4
Mato Grosso do Sul 2.8 Rio Grande do Sul 3.3
Alagoas 2.8 Rondônia 3.3
Sergipe 2.8 Tocantins 3.3
Pernambuco 2.7 Piauí 3.2
Amapá 2.7 Amapá 3.1
Bahia 2.7 Maranhão 3.1
Paraíba 2.6 Paraíba 3.1
Mato Grosso 2.6 Mato Grosso 3
Pará 2.6 Pará 3
Rio Grande do Norte 2.6 Bahia 2.9
Maranhão 2.4 Alagoas 2.8
Amazonas 2.3 Rio Grande do Norte 2.8
Piauí 2.3 Sergipe 2.6
2005 2015
To Influence Ideas: Brazil and Mexico launched CCTs in 1997
• Mexico implemented Progresa in a randomized rollout with rigorous evaluation • Brazil did no evaluation of Bolsa Escola
What are we learning in education?
• Big increase in rigorous evaluation over past 15 years
• Recent meta-studies
- Glewwe, 2014 - Evans and Popova, 2015 - Ganimian and Murnane, 2016
• 223 studies in 58 countries • Most comprehensive and nuanced framework • “Theory of action” approach
- Snilsveit et al, 2016 – evidence from low-income countries - JPAL (forthcoming) – compares cost-effectiveness
What can improve education results?
Children attend school and learn
Source: Ganimian & Murnane (2016)
Key goal:
What can improve education results?
Children attend school and learn
Build demand for education
Improve supply of education
Source: Ganimian & Murnane (2016)
What can improve education results?
Children attend school and learn
Build demand for education
Improve supply of education
Reduce the costs of attending school
1. Reduce the direct costs of schooling
2. Reduce the costs of complements
3. Improve school amenities
Prepare children for school
1. Provide children with vital medications
2. Improve nutrition
3. Improve parenting practices
Make schooling pay
1. Compensate families for opportunity costs
2. Inform families of the benefits of schooling
3. Offer incentives to students
Improve access to better schools
1. Provide information about school quality
2. Expand schooling options
Increase the quantity and quality of resources
1. Increase resources at school
2. Increase resources at home
3. Expand instructional time
Address students’ learning needs
1. Help teachers personalize instruction
2. Provide additional help to students
3. Lett students learn at their own pace
Increse teacher and/or principal effort
1. Increase the role of parents in school management
2. Offer incentives to teachers and principals
3. Hire teachers on fixed-term contracts
Build teachers’ capacity
1. Train teachers on specific skills
2. Prepare lessons for teachers
3. Let students teach each other
Source: Ganimian & Murnane (2016)
Reducing the direct costs of schooling
1 Lowering or eliminating school fees (Barrera-Osorio et al 2007; Gajigo 2014; Borkum 2012; Liu et al 2013).
2 Building new schools closer to students. (Duflo 2001; Mocan & Cannonier 2012; Burde & Linden 2013; Liu et al 2010; Berlinski et al 2009; Martinez et al 2013; Bouguen et al 2014).
3 Providing transportation to school. (Muralidharan & Prakash 2013).
Consistent evidence that these increase enrollments.
Reducing the complementary costs of attending school
Reducing the direct costs of schooling
1
Reducing the costs of complements to schooling
2
Reducing the costs of complements to schooling
1 Free school uniforms and eyeglasses (Evans et al 2009; Ma et al 2014; Glewwe et al 2014; Kremer et al 2003).
Evidence that these raise the learning achievement of students who receive them
What can improve education results?
Children attend school and learn
Build demand for education
Improve supply of education
Reduce the costs of attending school
1. Reduce the direct costs of schooling
2. Reduce the costs of complements
3. Improve school amenities
Prepare children for school
1. Provide children with vital medications
2. Improve nutrition
3. Improve parenting practices
Make schooling pay
1. Compensate families for opportunity costs
2. Inform families of the benefits of schooling
3. Offer incentives to students
Improve access to better schools
1. Provide information about school quality
2. Expand schooling options
Increase the quantity and quality of resources
1. Increase resources at school
2. Increase resources at home
3. Expand instructional time
Address students’ learning needs
1. Help teachers personalize instruction
2. Provide additional help to students
3. Lett students learn at their own pace
Increse teacher and/or principal effort
1. Increase the role of parents in school management
2. Offer incentives to teachers and principals
3. Hire teachers on fixed-term contracts
Build teachers’ capacity
1. Train teachers on specific skills
2. Prepare lessons for teachers
3. Let students teach each other
Source: Ganimian & Murnane (2016)
And even with all this… Geographical bias: 40% of evaluations from only 4 countries (China, India, Kenya, Mexico)
Topical bias: many cells have only 1 -2 country studies: 20% of total sample is CCT studies
Temporal bias: almost no studies measure impacts more than one year out
Non-comparable measures of learning: many evaluations use specialy designed tests which cannot be directly compared to tests used in other evaluations
Cost-effectiveness estimates limited: many evaluations do not present complete cost estimates and even good cost-estimates are limited by fact that the numerator (amount of learning increase) per dollar spent are not comparable
And even with all this…(most serious of all)
System reform “black hole”: hugely important reforms that are implemented system-wide are rarely evaluated
Curriculum reform
Mutli-shift schooling
Decentralization
Teacher career reforms: changes in the pay scale; standards for hiring; promotion criteria; performance evaluation; possibility of dismissal
Top 10 policies based on
evidence
Weighted by:
- Strength of evidence
- Cost-effectiveness/fiscal impact
- Generalizability
- Implementability
- Politics
“Don’ts” are as important as “Do’s”
Don’t raise teacher salaries without raising standards
Indonesia doubled teacher salaries in 2008
Only change in teacher performance expectations was a certification that all teachers could meet
Rigorous evaluation in 2014 (De Rhee et al.) found
No increase in student learning
No change in observable teacher effort
Huge fiscal deficit
Politically impossible to reverse
In Latin America, higher pay for teachers who pass tough competency exams – evidence this works (Mexico, Estrada 2013; Peru, 2014; Chile 2016)
Don’t expect cash transfers to raise learning
Cash transfers are effective in raising enrollments and student attendance in many countries
Size of these effects depend on whether:
Conditionality is monitored
Targeting is effective
Political winner, but impossible to reverse and costly over long-term
Best strategy to target carefully and monitor conditions
Complementary supply-side actions are needed for learning impact
Schultz 2004; Maluccio & Flores 2005; Galiani & McEwan 2013; Behrman et al 2009; Barrera-Osorio et al 2011; (Kremer et al 2009; Barrera-Osorio & Filmer 2015
Don’t expect computers to raise learning in the classroom
Expensive one-laptop-per child programs in Peru, Honduras and Colombia failed to increase student learning (Cristia et al. 2012; Barrera-Osorio and Linden, 2009)
Teachers failed to integrate them into classroom practice (classroom observations in Peru showed ITC use only 1% of class time)
BUT computer-assisted learning for targeted students in after school programs can be a powerful tool for remedial education
Banerjee et al 2007; Lai et al 2013; Mo et al 2014; Linden 2008; Lai et al 2015; Lai et al 2012; Yang et al 2013; Huang et al 2014.
Don’t expect parent involvement to transform school mgt.
Giving parents voice on school councils has generally had little impact on school results
Providing school grants that communities can manage has had some positive impacts
Niger, Mexico, El Salvador, Nicaragua – higher parent involvement in schools and satisfaction
The Gambia: lower student and teacher absence
But no strong impacts on learning anywhere
Slight exceptions: Senegal (small increase in girls’ test scores, but not boys) and Kenya (increased student test scores when combined with the hiring of an extra teacher)
Beasley & Huillery 2015; Gertler et al 2012; Blimpo, Evans and Lehire 2015; Carneiro et al 2015; Pradhan et al 2014; Lassibille et al 2010; Glewwe & Maïga 2011
Do build more schools
Decreasing the distance to school has big effects on student enrollment (.38 SD), attainment, and long-term wages
Also raises learning (.19 SD)
Improving school facilities (hygiene) also raises enrollment and attendance substantially
Big payoffs to:
Cost-effective construction
Careful mapping and planning
Duflo, 2001 evaluated impacts of Indonesia’s school expansion from 1973-78 (61,000 schools) using creative RD design
What about reducing class size?
Need evidence!
Do leverage the private sector Liberia’s Partnership Schools improved school management and
raised student learning 60% in one year
Teachers increased time on teaching from 32% to 48% (Stallings observations) and worked longer school day
Student attendance improved
BUT year 1 costs are unsustainable, so important to analyze further
Targeted vouchers for private schooling have:
Raised learning and attainment (Colombia)
Been cost-effective (India)
Romero, Sandefur and Sandholtz, 2017; Angrist et al 2002; Bettinger et al 2010; Angrist et al 2006; Muralidharan & Sundararaman 2015
Do invest in high quality student assessment
Learning assessment is a critical platform; makes it possible to:
Track system progress
Evaluate different policies
Reward schools for results
Compare private school quality and cost-effectiveness and make smart use of targeted vouchers
Inform parents about the benefits of schooling
Inform parents’ school choices
Empower parents to demand quality
A national student assessment system unlocks the door to
evidence-based policy
Do promote results focus and teacher interaction
School-based bonus pay can stimulate teachers to work together and raise student learning
Muralidharan & Sundararaman 2011; Muralidharan 2012; Lavy 2009, 2013; Contreras and Rau, 2012; Behrman et al, 2015; Ferraz and Bruns, 2014
Giving teachers feedback and coaching can also promote interaction and raise learning
Bruns, Costa and Cunha, 2017
Key conditions for impact:
Good quality national assessment that generates learning data for all schools (census-based)
Capacity to set school-level targets each year (or every 2 years)
Administrative integrity
Do equip teachers for multi-grade classes and big classes
Effective strategies for multi-grade classrooms include:
Self-paced learning materials
Students tutoring each other (cross-peer learning)
These can be helpful in large classrooms too
Organize classrooms so that all students get some individual attention and small group experience
Colombia’s Escuela Nueva an excellent model; Singapore’s student support policy Psacharopoulos, Rojas and Velez, 1992; Wachanga & Mwangi 2004; Beuermann et al 2013
Do prioritize early literacy and numeracy skills
“Structured pedagogy” and “Teaching to the right level” improve learning
Focus the curriculum for the first 3 grades on getting all children to read and comprehend
Assess students’ progress and group them by learning level, not age (EGRA, EGMA)
Provide remedial instruction (including computer-based programs and after-school teacher aides) to children who need it
Give teachers ongoing feedback and support
Liberia EGRA-plus Program – children scored .72 SD higher on reading
Mali Read-Learn-Read Program - +.24 SD
South Africa English and Operacy Program - + .40 SD
Ceara Brazil – “On Time Literacy” Program - +.17
Snilsveit, 2015; Piper and Korda, 2010; Costa and Carnoy, 2014; Banerjee et al 2007; Lavy & Schlosser 2005; Duc & Baulch 2012)
Conclusions
Huge increase in the evidence base in education is spreading promising ideas across the world
Replication is strengthening the evidence – but most consistent result is that country context, design details and implementation fidelity matter
Evaluation is expensive: should be focused on innovative programs and expensive policies
Where you can make a difference:
Thinking through the results chain/theory of action for new programs
Evaluating programs as they go to scale
Tracking learning impacts over the medium-term (4-5 years)
Finding creative ways to build evidence on key system-wide reforms
Esp, strategies for driving down class size
Thank you! [email protected]
Improving education in developing countries: Lessons from rigorous impact evaluations
Alejandro J. Ganimian Education Post-Doctoral Fellow, J-PAL
Outline 1. Demand: How to increase school participation 2. Supply: How to improve the quality of instruction 3. Action: How to use research to inform policy
Framework: How should we draw lessons?
Children attend school and learn
Build demand for education
Improve supply of education
Reduce the costs of attending school
1. Reduce the direct costs of schooling
2. Reduce the costs of complements
3. Improve school amenities
Prepare children for school
1. Provide children with vital medications
2. Improve nutrition
3. Improve parenting practices
Make schooling pay
1. Compensate families for opportunity costs
2. Inform families of the benefits of schooling
3. Offer incentives to students
Improve access to better schools
1. Provide information about school quality
2. Expand schooling options
Increase the quantity and quality of resources
1. Increase resources at school
2. Increase resources at home
3. Expand instructional time
Address students’ learning needs
1. Help teachers personalize instruction
2. Provide additional help to students
3. Lett students learn at their own pace
Increse teacher and/or principal effort
1. Increase the role of parents in school management
2. Offer incentives to teachers and principals
3. Hire teachers on fixed-term contracts
Build teachers’ capacity
1. Train teachers on specific skills
2. Prepare lessons for teachers
3. Let students teach each other
Source: Ganimian & Murnane (2016)
Providing children with vital medications
1 Deworming drugs—either alone or in combinations with micronutrients. (Miguel & Kremer 2004; Grigorenko et al 2006; Bobonis et al 2006; Jinabhai et al 2001; Maluccio et al 2009; Solon et al 2013).
Reduced absenteeism, but no consistent effects on learning.
2 Malaria drugs – (Brooker & Halliday 2015; Fernando et al 2006; Clarke et al 2008; Jukes et al 2006; Cutler et al 2007; Sylvia et al 2013).
No impact on student learning in short-or long-term.
Evidence: Preparing children for school
Improving nutrition
2
Providing children with vital medications
1
Improving nutrition
1 Iron supplements have generally improved children’s cognitive skills. (Baumgartner et al 2012; Manger et al 2008; Luo et al 2012; Rico et al 2006; Kleiman-Weiner et al 2013).
2 Other nutrients, such as fatty acids, have increased school participation, but not student achievement. (Hamazaki et al 2008; Mahawithanage et al. 2007; Osendarp et al 2007).
3 Free school meals – inconsistent impacts depending on the type of food and mode of delivery, among other factors. (Whaley et al 2003; Adelman et al 2008; McEwan 2013).
Preparing children for school
Improving parental practices
3
Improving nutrition
2
Providing children with vital medications
1
Improving parenting
1 Home visits to teach mothers how to stimulate young children’s cognitive and psycho-social skills. (Attanasio et al 2014; Walker et al 2005; Tieffenberg et al 2000; Rocha & Soares 2010; Kagitcibasi et al 2001; Rosero & Oosterbeek 2011).
Positive impacts on school attendance, cognitive skills, and adult wages.
Compensating parents for opportunity costs
1 Conditional cash transfers . (Schultz 2004; Maluccio & Flores 2005; Galiani & McEwan 2013; Behrman et al 2009; Barrera-Osorio et al 2011). (Kremer et al 2009; Barrera-Osorio & Filmer 2015).
Consistent positive impacts on enrollment and
attendance but not learning. Size of effects depends on design.
Evidence: Making schooling pay
Compensating parents for foregone opportunities
1
Informing families of the benefits of schooling
2
Informing families of the benefits of schooling
1
2
. (Jensen 2010, 2012).
2 But mode of delivery of the information makes a difference. (Nguyen 2009).
Give students and parents information on the
returns to schooling (e.g., the wages of
individuals with different levels of schooling)
Positive impacts on attainment and learning.
Evidence: Making schooling pay
Compensating parents for foregone opportunities
1
Informing families of the long-term benefits of schooling
2
Offering incentives
to students 3
Offering incentives to students
1 Rewarding students for test scores has not produced consistently positive results and in some cases has stimulated cheating. (Angrist & Lavy 2009). Sharma 2010; Berry 2014; Li et al 2014). Behrman et al 2015).
Providing information about school quality
1 Providing parents with information on schools’ performance may help them make better decisions about where to send their children to school and to demand changes in their schools.
2 The evidence is modest, but encouraging. The increased learning has mostly been in private, not public schools. This may be because they have more flexibility to respond to information. (Andrabi et al 2009; Mizala & Urquiola 2013; Camargo et al 2011).
Evidence: Improving access to better schools
Providing information about school quality
1
Expanding schooling
options 2
Evidence: Expanding schooling options
1 “Universal” voucher programs are difficult to evaluate, but appear to have little effect on overall system learning results and attainment, but do increase student sorting. (Hsieh & Urquiola 2006; Contreras 2001).
2 “Targeted” voucher programs giving low-income parents money to send children to private schools have improved enrollment, attainment, and achievement. (Angrist et al 2002; Bettinger et al 2010; Angrist et al 2006).
3 Voucher winners in private schools performed no better than students who stayed in public schools, but private schools operated at a third of the cost of public schools. (Muralidharan & Sundararaman 2015).
Increasing resources at school
1 Providing schools with inputs such as textbooks, libraries, and flipcharts has typically not improved student achievement.
2 Sometimes, they are not even used. (Glewwe et al 2009; Sabarwal et al 2013).
3 Other times, they do not improve children’s daily experiences—typically, because teachers lack the knowledge to use these resources to improve instruction. (Glewwe et al 2004; Borkum et al 2013; Barrera-Osorio & Linden 2009; Cristia et al 2012).
4 Yet other times, school resources “crowd out” resources at home. (Das et al 2013; Björkman 2007).
Evidence: Increasing the quantity and quality of resources
Increasing resources at home
2
Increasing resources at school
1
Evidence: Increasing resources at home
1 Many children in developing countries lack basic educational resources at home. To stimulate learning, several initiatives have provided children with resources—typically, computers.
2 These initiatives have not improved student achievement. In fact, students report using computers for playing games. (Malamud & Pop-Eleches 2011).
3 In some contexts, students get more apt at handing the computer that they are given, but not at computer skills. (Beuermann et al 2013).
Evidence: Increasing the quantity and quality of resources
Expanding instructional
time
3
Increasing resources at home
2
Increasing resources at school
1
Evidence: Expanding instructional time
1 There is only one study of the effects of simply increasing instructional time. This study finds an impact on student learning. (Orkin 2013).
2 Other evaluations of initiatives that lengthened the school day have also captured the effects of simultaneous reforms (e.g., more resources, infrastructure upgrades, and increased per-pupil spending). (Bellei 2009).
Evidence: Helping teachers personalize instruction
1 Class size reductions, by themselves, have not improved student learning. Yet, they have improved test scores when students were “tracked” into different classrooms based on their performance on an initial assessment. (Duflo et al 2007, 2011).
2 Tracking students across schools has improved achievement, but also elicited unintended consequences (e.g., teacher sorting). (Pop-Eleches & Urquiola 2013; de Hoop 2010).
3 Providing teachers with feedback on students’ learning has improved results only when accompanied by actions to build teacher capacity or make their jobs manageable. (Muralidharan & Sundararaman 2010; Duflo et al 2015; Piper & Korda 2011).
Evidence: Providing additional help to students
Helping teachers
personalize instruction
1
2 Providing additional help to students
Evidence: Providing additional help to students
1 Low-achieving students often get very little from regular instruction. Pulling them out of the classroom to provide them with remedial instruction has improved their performance. (Banerjee et al 2007; Lavy & Schlosser 2005; Duc & Baulch 2012).
2 The merits of non-academic remedial support, however, are more uncertain. (Cabezas et al 2011; Huan et al 2014).
Evidence: Addressing students’ learning needs
Helping teachers
personalize instruction
1
2 Providing additional help to students
Letting students learn at
their own pace
3
Evidence: Letting students learn at their own pace
1 Many initiatives have experimented with computer-assisted learning (i.e., questions that adjust to students’ skill level). These programs have generally improved student achievement. (Banerjee et al 2007; Lai et al 2013; Mo et al 2014).
2 These initiatives are most effective when used as a complement to, rather than a substitute for, regular classroom instruction. (Linden 2008; Lai et al 2015; Lai et al 2012; Yang et al 2013; Huang et al 2014).
2 These programs are often combined with free laptops, which by themselves have no impact on student achievement. (Mo et al 2013; Carrillo et al 2011).
Evidence: Increasing teacher and/or principal effort
Increasing the role of parents in school management
1
Evidence: Increasing the role of parents in school management
1 Many interventions have provided information to communities about how to get involved in school management. Yet, more information has not always translated into more action. (Banerjee et al 2010; Barr et al 2012).
2 Some interventions have affected intermediate outcomes, but have had small and isolated effects on student achievement. (Pandey et al 2009).
3 Providing schools with monetary grants that communities can manage have varied widely in their effects, depending on the pre-existing level of community capacity. (Beasley & Huillery 2015; Gertler et al 2012; Blimpo & Evans 2011; Carneiro et al 2015; Pradhan et al 2014; Lassibille et al 2010; Glewwe & Maïga 2011).
Evidence: Increasing teacher and/or principal effort
Increasing the role of parents in school management
1
2 Offering incentives to teachers/principals
Incentives to teachers/principals
1 Doubling teacher salaries without increasing standards has no impact on teacher effort or student achievement. (de Ree et al 2015)
2 Rewarding teachers’ for effort (i.e., to attend school) has increased their effort and student learning when it has relied on non-discretional, systematic monitoring. (Duflo et al 2012; Chen et al 2001; de Laat et al 2008).
3 Rewarding teachers’ for performance (i.e., to increase test scores) has increased student learning in the short- and medium-term. (Muralidharan & Sundararaman 2011; Muralidharan 2012; Lavy 2009, 2013).
4 In some contexts, these programs elicited “teaching to the test”. (Glewwe et al 2010).
Evidence: Increasing teacher and/or principal effort
Increasing the role of parents in school management
1
2 Offering incentives to teachers/principals
3
Hiring teachers on fixed-term contracts
Evidence: Hiring teachers on fixed-term contracts
1 In many countries, it is very difficult to dismiss public school teachers for poor performance. This has led some to hire teachers on fixed-term, renewable contracts.
2 In the short-term, these initiatives have resulted in higher effort among “contract” teachers and higher student achievement. (Muralidharan & Sundararaman 2013).
3 Yet, in some contexts, contracts were given to relatives of existing teachers and regular teachers reduced their effort. (Duflo et al 2015).
4 These initiatives have been more difficult to implement when led by governments, resulting in no gains in teacher effort or learning. (Bold et al 2013).
Evidence: Training teachers on specific skills
1 Training teachers “on-the-job” is difficult. The few programs that have been evaluated have documented wide variability in program implementation and no effect on student learning. (Yoshikawa et al 2015; Angrist & Lavy 2001).
2 However, even when faithfully implemented, these programs have had no effect on student achievement. (Zhang et al 2013; Yue et al 2012).
Evidence: Building teachers’ capacity
1 Training teachers
on specific skills
2
Preparing lessons for teachers
Evidence: Preparing lessons for teachers
1 Countries where teacher capacity is low have experimented with “scaffolding” in which teachers are given the materials to teach a lesson and trained on how to teach it.
2 Some prepare lessons for teachers to facilitate students’ independent learning (e.g., a “readathon”). (Abeberese et al 2014).
3 Others prepare lessons that teach teachers how to impart a specific skill (e.g., early grade reading). (Lucas et al 2014; Githua & Nyabwa 2008; Spratt et al 2013; Van Staden 2011; He et al 2008).
4 In some cases, the lesson is actually scripted or pre-recorded. (Naslund-Hadley et al 2014; Piper 2009).
Evidence: Building teachers’ capacity
3
Letting students
teach each other
1 Training teachers
on specific skills
2
Preparing lessons for teachers
Evidence: Letting students teach each other
1 Some initiatives have attempted to bypass low teacher capacity by encouraging peer-to-peer learning.
2 In secondary school, when students are more independent, these programs have had positive impacts on student learning. (Wachanga & Mwangi 2004; Beuermann et al 2013).
3 In primary school, when students rely on the teacher as facilitator, these programs have been far less successful. (Berlinski & Busso 2013).
Evidence: Important caveats in interpreting current evidence
1 Details of design and implementation matter.
2
Average effects may mask considerable heterogeneity by groups.
3
Consequences are likely to depend on circumstances in the particular setting. • Many children out of school → focus on demand-side
interventions. • Most children in school, low learning → improving
instruction is crucial. • Teachers have capacity but low effort → aligning the
incentives of students, parents, and teachers may make more sense.
4 Most of what we know concerns short-term outcomes. Examining longer term consequences of promising interventions is important.
Thank you E-mail: [email protected] Website: http://scholar.harvard.edu/alejandro_ganimian