Aspirations, Stereotypes and Educational Choices Eliana La Ferrara - - PowerPoint PPT Presentation

Aspirations, Stereotypes and Educational Choices

Eliana La Ferrara

Bocconi University, IGIER and LEAP DEC Lecture The World Bank – May 6, 2019

Motivation Aspirations and poverty

• Standard poverty traps are generated by external constraints:

malnutrition (Dasgupta-Ray 1986), credit mkt imperfections (Banerjee-Newman 1991), etc.

• Internal constraints may also generate poverty traps, leading to

aspiration failures (Appadurai 2004; Dalton, Ghosal, Mani 2014; Genicot-Ray 2016) Poverty  low asp.  low investment  poverty...

Stereotypes and poverty

• Above mechanism particularly relevant for groups that are

negatively stereotyped or discriminated in society. Stereotype threat: anticipate & internalize stereotype  underperform (Steele-Aronson 1995)

This lecture

• How do aspirations and stereotypes play out in a context where

investment has long term consequences, i.e., education?

• Study two different potentially marginalized groups: immigrants

in Italy and Black South-Africans in SA

• Combine administrative & experimental data
• Consider different social interactions: w/ teachers & w/ peers

Plan of the talk

1. Educational choices of immigrant children in Italy

• Aspirations & psychological factors

2. Teachers’ stereotypes

• Effect of revealing own bias

3. Peers’ stereotypes

• Effect of inter-group contact

• 1. Goals and gaps:

Educational Careers of Immigrant Children

(with M. Carlana and P. Pinotti)

• In most countries, children of immigrants disadvantaged in labor

mkt due to lower educational attainment

10 20 30 40 50 60

Males

natives immigrants

5 10 15 20 25 30 35 40

Females

natives immigrants

% of aged 20-29 w/ low attainment

Source: OECD, 2011

Early tracking and educational choices

Carlana, La Ferrara, Pinotti Goals and gaps

Greater disadvantage in educational systems characterized by early tracking (Cobb-Clark et al. 2009)

• Information gaps (Dustmann et al. 2014)
• Aspiration traps (Genicot and Ray, 2014; Guyon and Huillery, 2014)

High-school choice is an early career decision w/ long term consequences on labor mkt outcomes (Giustinelli, 2011)

Research questions

• Do immigrant children under-aspire when choosing high

school in an early tracking system?

• Can we align their aspirations w/ their academic potential?
• Evaluate a large-scale program targeting high-achieving

immigrant students in Italy

The Italian schooling system

Stratification of students after 8th grade into 3 tracks

• Academic oriented (liceo)  college
• Technical  college or white collar jobs
• Vocational  blue collar jobs

Carlana, La Ferrara, Pinotti Goals and gaps

Vocational track worse under several dimensions (e.g., earnings, ex post satisfaction, etc.)

Fact 1: Educational segregation

• Immigrant students disproportionately choose vocational track

compared to Italians

Carlana, La Ferrara, Pinotti Goals and gaps

• Need to control for ability  use standardized test score in

math & italian (Invalsi) in 6th grade as proxy for academic potential

Track choice by quintile of Invalsi score

Carlana, La Ferrara, Pinotti Goals and gaps

.2 .4 .6 .8 1 1 2 3 4 5 Test score, 6th grade, quintiles Italians Immigrants

Males

Probability of enrolling in “high track” (Liceo/Technical) Immigrant boys “under-aspire” compared to native boys

Track choice by quintile of Invalsi score

Carlana, La Ferrara, Pinotti Goals and gaps

Probability of enrolling in “high track” (Liceo/Technical) No segregation for girls at quintiles 3-5

.2 .4 .6 .8 1 1 2 3 4 5 Test score, 6th grade, quintiles Italians Immigrants

Females

Teachers’ recommendations

Carlana, La Ferrara, Pinotti Goals and gaps

.2 .4 .6 .8 1 1 2 3 4 5 Invalsi score, 6th grade, quintiles Italians Immigrants

Males

.2 .4 .6 .8 1 1 2 3 4 5 Invalsi score, 6th grade, quintiles Italians Immigrants

Females

Teachers' Recommendation (Liceo or Technical)

Teachers less likely to recommend high track to immigrants, conditional on academic performance

The intervention

“Equality of Opportunity for Immigrant students” (EOP)

• Program in collaboration w/ Ministry of Education (MIUR)
• Target: high-performing students from low-income countries in

lower secondary school

• Goal: align their HS choice w/ their academic potential

Schools

• All schools w/ >20 immigrant students in 5 provinces of

Northern Italy  145 schools: randomize 70 treatment, 75 control

Carlana, La Ferrara, Pinotti Goals and gaps

Individuals

• In each school, the 10 immigrant students w/ highest Invalsi

test score in 6th grade (only countries w/ GDP pc < Italy)  Takeup rate: 79%

The intervention

Students followed during grades 7 and 8. Two types of activities:

Carlana, La Ferrara, Pinotti Goals and gaps

• 1. Career choice consultancy
• Information about Italian schooling system
• type of high-schools, job opportunities, booklet translated in

language of home country

• Psychological support based on Social Cognitive Career Theory
• 14 meetings during grades 7-8: 5 group meetings, 5 individual,

3 w/ parents, 1 w/ teachers

Examples of psychological support activities

Carlana, La Ferrara, Pinotti Goals and gaps

• Peer education, e.g., video on barriers and self-efficacy of high-

school students

• “Thinking about your past life, indicate 5 study experiences and

5 other experiences that you have completed successfully… which personal resources helped you doing well in that thing”

• “Please find below the professions you selected and indicate

which resources are needed (knowledge, skills, personality traits, motivations, ... ) then divide them into “I have it” and “I need to develop it”

• “Please list the results you would like to achieve with your job”

The intervention

Carlana, La Ferrara, Pinotti Goals and gaps

• 2. Cognitive Academic Language Proficiency (CALP)
• Tutor on Italian language to facilitate studying and learning all

subjects

• # meetings higher for students w/ lower Invalsi scores in

grade 6 (2 thresholds, though little variation)

• The 2 components (Career consultancy & CALP) offered as joint

package, not a 2x2 design

• “ethics”

Data

Carlana, La Ferrara, Pinotti Goals and gaps

• 1. Ministry of Education, University and Research (MIUR):

information on educational career (enrollment, failure rates, teachers’ recommendations, final grades)

• 2. Italian Agency for the Evaluation of Educational System

(INVALSI): standardized test scores in grade 6 and 8, information on family background

• 3. First-hand data: questionnaire on psychological traits

(academic motivation, perception of economic and social barriers on work and educational career)  Balance on entry level score & socio-economic characteristics

Impact: High school choice

Carlana, La Ferrara, Pinotti Goals and gaps

11.8% increase

• ver the mean

Impact: High school choice

Carlana, La Ferrara, Pinotti Goals and gaps

Closing the gap w/ natives

Note: comparable Italian students matched on 6th grade INVALSI score

.65 .7 .75 .8 .85 Control Treat Italians

Males

.65 .7 .75 .8 .85 Control Treat Italians

Females

Probability of high-track at the end of lower secondary

Impact: Grade retention

Carlana, La Ferrara, Pinotti Goals and gaps

Failure 7th or 8th grade

Potential mechanisms

Carlana, La Ferrara, Pinotti Goals and gaps

• 1. Cognitive skills: Invalsi score in Italian & Math at the end of

grade 8

• 2. Non-cognitive skills: questionnaire on psychological traits
• 3. Teachers’ recommendations on HS track

• 1. Cognitive skills improve for boys

Carlana, La Ferrara, Pinotti Goals and gaps

• 1. Cognitive skills improve for boys

Is it because of academic tutoring?

• # of meetings for Italian language tutoring (CALP) was a

function of Invalsi 6th grade score Z  RDD around threshold of 65

Carlana, La Ferrara, Pinotti Goals and gaps

Students in sample # meetings 7th grade # meetings 8th grade

Z ≤ 65 64% 22 33 65 < Z ≤ 80 32%

• 33

Z > 80 4%

Impact of increased # meetings on test scores

Carlana, La Ferrara, Pinotti Goals and gaps

Math Italian

No significant effect

• 2. Non-cognitive skills

Questionnaire on a broad range of psychological variables, including info on:

• Goals: e.g., which educational level do you want to achieve?
• Self-efficacy: e.g., thinking about your ability, do you think you

could get a university degree?

• Perceived barriers: e.g., do you think the following barriers

could be an obstacle to achieving your educational goals: economic barriers, racial prejudice, ideas of the family…?

Carlana, La Ferrara, Pinotti Goals and gaps

Aggregate into 2 indexes through principal component analysis:

• Academic motivation
• Perceived barriers

Treatment improves non-cognitive skills

Carlana, La Ferrara, Pinotti Goals and gaps

• 3. Teachers’ recommendations

Probability of recommending high track

Carlana, La Ferrara, Pinotti Goals and gaps

Effect only for boys, who had shown improvements in academic performance  not driven by experimenter demand effects

Additional results

Variance decomposition

• Most of the effect explained by motivation & teachers’

recommendation

Carlana, La Ferrara, Pinotti Goals and gaps

Medium term effects

• 2 years later, treated students are no more likely to fail subjects,

despite attending more demanding high schools Spillovers

• Male classmates of treated students have lower retention rates;

impact stronger on immigrants

• Improvement in “behavior” of immigrant males
• No impact on high school choice for boys, but (+) impact on

female classmates (who were closer to “marginal”)

• No impact on teachers’ recommendations

Summing up

• EOP reduced educational segregation: +12% enrollment in

demanding high schools for immigrant boys, closing gap w/ comparable Italians

• Mechanisms: motivation & teachers’ support

Policy implications

• Role of ‘soft skills’ suggests potential information &

“aspirations” interventions, ideally scalable & cheaper to implement

Carlana, La Ferrara, Pinotti Goals and gaps

• Role of teachers’ recommentations suggests working on

teachers’ attitudes towards immigrant students, in particular stereotypes  next paper

• 2. Revealing stereotypes:

Evidence from immigrants in schools

(with A. Alesina, M. Carlana and P. Pinotti)

Stereotypes are over-generalized representations of differences between groups (Bordalo et al., 2017)

• Allow for easier and efficient processing of information
• May cause discrimination, i.e. biased judgment against

particular groups If discrimination occurs in a critical period for educational choices, this may discourage investment by disadvantaged groups

• Important to study negative stereotypes in school
• Link b/w stereotypes and aspirations

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Research questions

Are teachers biased against immigrants?

• Analyze the impact of teachers’ stereotypes on bias in grading

(i.e., the difference b/w teacher-assigned grades & blindly graded test scores) How do we change teachers’ bias?

• Evaluate a simple & scalable intervention: reveal own

stereotypes to teachers

• Test if this affects their grading policy
• Indirect evidence about people’s awareness of their own bias

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Data

• 102 middle schools in Northern Italy
• 1.384 math and literature teachers
• Middle school: grades 6-8, same teachers & classmates for 3

years

Teacher survey

• IAT, demographic information, explicit beliefs & attitudes

towards immigrants

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Student data from 2 admin. sources:

• Italian Ministry of Education (MIUR): teacher-assigned grades in

math and literature

• National Evaluation Agency (INVALSI): standardized test scores

(blindly graded), family background

Implicit Association Test (IAT)

• Experimental method from social psychology (Greenwald and

Banaji, 1995)

• Pair two concepts in rapid categorization task
• Speed in associating: mental process perceives a given pair as

less common

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Advantages of IAT

• Reveal cognitive processes of which individuals may not be

aware (e.g., perception, stereotyping)

• Or may be uncomfortable disclosing (e.g., prejudice)

IAT: Example

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Distribution of the IAT in our sample

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Correlates of IAT

IAT correlated w/ explicit attitudes

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Correlates of IAT

IAT uncorrel. w/ past performance

• f natives vs

immigrants

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Results I: Bias in grading

Teacher-assigned vs. blindly graded test scores

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Interpreting the grade gap

• Existing literature takes grade gap as evidence of discrimination

(Gilliam et al., 2016; Botelho et al., 2015; Burgess and Greaves, 2013; Hanna and Linden, 2012; Van Ewijk, 2011; 2018; Lavy, 2008)

• However, it could reflect unobservables:
• e.g., multiple choices standardized test scores may be easier

for immigrants than teacher-assigned exams

• e.g., teachers may observe characteristics (effort, discipline)

that differ b/w natives and immigrants To isolate bias, we test if the gap in grading is correlated w/ teachers’ IAT

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Bias in grading of Math teachers

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Bias in grading of Math teachers

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

No correlation w/ IAT for Literature teachers

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Why different results for Math & Literature teachers?

• Standardized test scores in math may be better at measuring

skills valued by teachers when grading than those in reading (Bettinger, 2012)

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

• Literature grades may reflect different standards for teachers

w/ different expectations on immigrants' language skills - and these expectations may be correlated w/ IAT

• We find that Literature teachers w/ higher IAT give better

grades to 1st vs 2nd generation immigrants, ceteris paribus

The experiment

• We offered all teachers the possibility of receiving an email w/

feeedback on their own IAT

• 80% asked to receive feedback
• Decision uncorrelated w/ IAT & other observables
• We estimate ITT

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

• Text of email:
• Brief description of what IAT does
• Placement into "slight", "moderate" or "strong", based on

Greenwald et al. (2009)

Timing of experiment

• We randomize the timing of feedback at school level:

2 weeks before vs. 2 weeks after the end-of-semester grading

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

 Observables balanced at teacher & student level

Impact of treatment

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Treated teachers increase grades of immigrants and decrease those

• f natives

Impact of treatment

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Literature teachers also adjust (but recall that they received “equally negative” feedback as math teachers)

ITT on teacher-assigned grades

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Testing information channel

Do teachers react b/c they were unaware of their own stereotypes?

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Approach 1: exploit information on explicit bias

• WVS question on right to job: teachers who openly state that natives

have more rights than immigrants should be less “surprised” by email feedback

• We expect smaller(or no) adjustment in grades

Explicit bias & impact of treatment

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Testing information channel

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Approach 2: Precision of the signal

• Our email provided feedback on 2 different IATs: one for male names

& one for female names:

• Some teachers got “strong/strong”, others got mixed

 expect stronger effect from more precise signal (i.e., both IATs “strong”)

Treatment effect by precision of the signal

Math teachers

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Literature teachers

Summary & policy implications

• Revealing own stereotypes to teachers reduces bias in grading
• Suggests that teachers were not aware of (or had not fully

internalized) their own stereotypes

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

• Interventions aimed at increasing awareness of implicit racial

stereotypes can help counteract discrimination

• e.g., committee members taking IAT

Implicit bias training: example 1

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

Implicit bias training: example 2

Alesina, Carlana, La Ferrara, Pinotti Revealing stereotypes

• 3. Interaction, stereotypes and performance:

Evidence from South Africa

(with J. Burns and L. Corno)

Corno, La Ferrara, Burns

Interaction, stereotypes & performance

Stereotypes not only by teachers but also by peers

• Can stereotypes be changed through interaction w/ members of

a different group?

• How does this affect academic performance?

Answer these questions exploiting random exposure to roommate in some dorms at University of Cape Town

• South Africa is a particularly interesting context, as apartheid led

to stereotyping & marginalization of blacks

Conceptual framework

Contact hypothesis (Allport, 1954)

• Interaction w/ other group  reduction of negative

stereotypes under certain conditions (equal status, common goals, gains from interdependence, authorities support inter-group contact)

Corno, La Ferrara, Burns Interaction, stereotypes & performance

“Negative contact” hp. (Paolini et al. 2010, Barlow et al. 2012)

• Heightened salience of difference in preferences

Institutional context & data

UCT residence allocation policy

• Assignment to residences is random
• Allocation to rooms within residence (single or double) done

by Warden

• 8 residences randomize room assignment  our sample

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Sample

• Freshmen who joined UCT in 2012 & live in double rooms
• Recruited for survey on “student life” at UCT

2 rounds of data

• Round 1 (Feb. 2012)
• 637 freshmen: 70% of universe in double rooms
• Round 2 (Sept. 2012)
• 517 out of 637  21% attrition. Attrition uncorrelated w/

treatment (mixed room), w/ baseline IAT & w/ interactions

IAT

Corno, La Ferrara, Burns Interaction, stereotypes & performance

We run 2 types of IAT

• “Population”: positive & negative attributes

 racial prejudice

• “Academic”: match pictures w/ pctiles of grade distrib.

 priors on academic benefits from interaction

Population IAT at baseline

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Whites vs Blacks

.2 .4 .6 .8 kdensity

• 2
• 1

1 Population IAT Whites Blacks

Negative values: negative stereotypes on blacks vs whites

Academic IAT at baseline

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Whites vs Blacks

.2 .4 .6 .8 1 kdensity

• 2
• 1

1 2 Academic IAT Whites Blacks

Negative values: blacks worse academic performance than whites

Empirical strategy

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Regress outcomes of interest on “MixRoom”: dummy =1 if roommate of different race. Treatment exogeneity

• Prob. of being in mixed room is

 uncorrelated w/ individual IAT, admission score, etc.  uncorrelated w/ difference between pair (dyadic)

Results I: Impact on stereotypes

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Treatment closes the gap in Popul IAT b/w whites and blacks

Results I: Impact on stereotypes

Corno, La Ferrara, Burns Interaction, stereotypes & performance

No impact on avg. on Academic IAT

• We find an impact when matched w/ high performing roommate
• f opposite race

Results II: Impact on academic performance

Corno, La Ferrara, Burns Interaction, stereotypes & performance Magnitude: .26 std dev. Closes 1/3 of the gap b/w blacks & whites

Results II: Impact on academic performance

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Effect is mediated by roommate’s stereotypes

Corno, La Ferrara, Burns Interaction, stereotypes & performance Move white roommate’s IAT from -.36 to 0  +.26 std dev GPA black

Discussion

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Why does black students’ performance improve?

• Not because (white) roommate is more skilled: we control for

admission score Possible channels

• Role modeling: learning how to navigate the system
• Network effects: different set of peers
• Fewer opportunities for joint “distractions”
• Anxiety reduction
• Impact on academic outcomes persists at the end of year 2 (not

for GPA but for # exams passed and eligibility to continue)

Results III: Attitudinal outcomes

Corno, La Ferrara, Burns Interaction, stereotypes & performance

20 outcome measures grouped in 3 indices

• 1. Friendships
• %friends and study mates of diff. race (actual & ideal),

how often hang out w/ diff. race

• 2. Attitudes
• Talk about race, affirmative action, dancing/dating other

group

• 3. Pro-social behavior
• Volunteer, money to charity, cooperate in prisoner’s

dilemma

Friendships

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Attitudes

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Pro-social behavior

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Summary & policy implications

Exposure to roommate of different race

• Changes stereotypes: reduction in negative stereotypes held by

whites

• Affects academic performance: blacks w/ non-black roommate

improve their GPA & take more exams

• Effect stronger if roommate less prejudiced

Corno, La Ferrara, Burns Interaction, stereotypes & performance

Implications for integration policies

• Not only effective in reducing negative stereotypes towards
• utgroup (contact hp.)
• But also functional to improving productivity/performance

THE END

Thank you!