vakis sept 9 oxford
TRANSCRIPT
Teacher biases, student attitudes and improving educational outcomes in
Peru
Renos Vakis
September 9, 2016
Behavioral Initiatives
Attitudes and world view (mental models)• Chapter 3, The World Development Report 2015
• They influence the process of decision making…and hence outcomes• Is there scope for policy?
Our biaseshttps://implicit.harvard.edu/implicit/takeatest.html
Stereotypes, mental models and achievement
Identidad asiatica Control Genero65
70
75
80
85
90
95
Math score(10 year old American girls from Asian
descent)Sc
ore
Asian-American Control Female
Nalini Ambady et al (2001)
Let’s go to Peru
Peru contextOngoing work in Ministry of Education exploring
1. Teacher attitudes and motivation2. Student attitudes and low performance
Research project 1: Do teachers use their biases to assess students’ performance?with Gabriela Farfan and Alaka Holla (World Bank)
• Teachers are key input for student’s performance
• Teacher expectations for their students affects school performance and student aspirations (Rosenthal y Jacobson, 1968, Upadyaya and Eccles, 2014)
• Teachers influence student’s long term outcomes (Chatty et al. 2013)• Non-observable teacher characteristics explain a large share of student performance
• How do teachers assess students’ performance?• We typically think of exams as objective indications of scholastic aptitude• But teachers often have some discretion in grading• Plus, kids are not always consistent so it might be hard to extract a clear signal from their
behavior
What we do• What we do: replicate and expand evidence by Darley and Gross (Journal
of Personality and Social Psychology, 1983) for teachers in Lima
• 600 teachers from 100 schools in Lima
• Teachers watch a video of a 9-year old student (Diego) taking an exam
• Random teacher assignments in one of two versions of background video• Diego comes from 2nd quintile (poor)• Diego comes from 3rd quintile (non-poor)
• Teachers assess Diego’s performance on three dimensions• Cognitive• Behavioral• Schooling expectations
Video timeline1.
Project presentati
on
2. Background video on Diego
3. Testing video
4. Survey on
Diego’s performan
ceDiego Pobre
Diego No Pobre
What happened?School grade
Poor Non-poor3
3.5
3.2
3.4
Poor Non-poor
Grad
e le
vel
Poor Non-poor0
20
40
60
40
60
Poor Non-poor
%
Expect to go beyond secondary
2 months20 pp
So…• Teacher systematically use their mental model to affect
their judgement and assess student performance
• By a lot
• And form a judgment on various domains (cognitive, expectations)
Areas for solutions
Education … information, awareness, alternative modelsExposure … social contact, role models, environmentApproach … higher level processing, reduce cognitive load, checklist, procedures (slow down)
STATUS: Debriefing, exploring….
Project 2: improving student motivation and educational outcomeswith Ingo Outes (Oxford) and Alan Sanchez (GRADE)
• Student school performance and aspirations can be affected by
• Parents• Self-identification into groups (math example and stereotypes,
also Neuville y Croizet, 2007; Steele & Aronson, 1995 )• Teacher expectations and beliefs (Rosenthal y Jacobson, 1968)
• Brain malleability and mental models
• Evolving evidence on traits and malleability (intelligence and personality)
• Changing mental models of brain malleability and link to improved motivation, effort and school outcomes ((Dweck et al. 1995; Dweck, 2006; Blackwell et al. 2007; Paunesku et al. 2012; Yeager et al. 2012)
05
1015
20C
hild
ren'
s E
duca
tiona
l Asp
iratio
ns
0 5 10 15 20Parent's Educational Aspirations
Children-Parent aspiration 45%
Children’s educational aspirations closely mirror those of their parents
What did we do?• Inspired by growth mindset literature to design a simple
intervention• Where
• 800 secondary schools in poorer neighborhoods in Lima, Ancash and Junin• Around 50,000 students in 1st and 2nd grade
• What• Introduced a 90 minute session on brain malleability• In-class assignment – write a letter to give advice on how to exercise brain• Class posters for salience• Low cost
• 2-3 months before end of school year• Use national test exam to evaluate test scores in math and
language
Overall impacts
Math (*) Spanish Index (*)-0.05
0.00
0.05
0.10
0.15
0.20
0.25ITT impact
ITT
Impa
ct%
of s
tand
ard
devi
-at
ions
School level, N = 799
Impacts by region
Math Spanish Index Math (**) Spanish Index (*)Lima, N = 391 Ancash-Junin, N = 408
-0.15
-0.05
0.05
0.15
0.25
0.35
0.45ITT impact
ITT
Impa
ct%
of s
tand
ard
devi
-at
ions
Impacts by quartile
Q1N =
4413
Q2N =
9033
Q3 (**)N =
4345
Q4N =
3909
Q1N =
4413
Q2N =
9033
Q3 (*)N =
4344
Q4 (**)N =
3910Math Spanish
-0.20
-0.10
0.00
0.10
0.20
0.30
0.40
0.50 ITT impact
ITT
Impa
ct%
of s
tand
ard
devi-
atio
ns
Summary• Large increases in test performance
• Extremely effective and low cost
• Hypothesis for mechanism: student AND teacher change of mental models
• Heterogeneous effects
Next steps• Phase 1
• Exploring mechanisms and heterogeneity of impacts further• Are impacts sustainable a year later?
• Phase 2 (adapting in 2800 schools/150,000 students) isolating teacher v. student channels
• Video vs Text• Single-Session vs Double-Session• Teacher Module• Timing of Delivery
• Adapting and replicating in other countries (Indonesia, Ghana…)