Jere R. Behrman
University of Pennsylvania
World Bank Conference
"Equity, Development and Policy:
Evidence, New Ideas & Future Directions"
10 June 2011

ECD has multiple interacting components:
Cognitive-language skills, Interpersonal skills,
Social-emotional skills and Physical development
(nutrition, sensory-motor)
Large numbers of children in developing world, estimated 200+ million, fail to reach their development potential -- four major risks: stunting, iron deficiency, iodine deficiency, and lack of cognitive and social-emotional stimulation*
*Grantham-McGregor, SM., YB Cheung, S Cueto, P Glewwe, LM Richter, BJ
Strupp, 2007, “Over Two Hundred Million Children Fail to Reach Their
Developmental Potential in the First Five Years in Developing Countries,”
Lancet 369 (January): 60–70.

Mean Z-scores for height-for-age relative to the new WHO standards for Peru, 2000
Z
-0.5
-1
0
0-5
Age (months)
6-11 12-17 18-23 24-29 30-35 36-41 42-47 48-53 54-59
-1.5
-2

(Engle et al 2011 Lancet)

(Engle et al 2011 Lancet)

What interventions have worked at scale to reduce the loss of developmental potential from the four major risks: stunting, iron deficiency, iodine deficiency, and lack of cognitive and social-emotional stimulation?
• Engle, Patrice L., Maureen M. Black, Jere R. Behrman, Meena Cabral de Mello, Paul
J. Gertler, Lydia Kapiriri, Reynaldo Martorell, Mary Eming Young, International Child
Development Steering Committee, 2007, “Strategies to Avoid the Loss of Potential
Among 240 Million Children in the Developing World,” Lancet 369 (January), 229-
242.

Type of intervention
Significant evaluations out of total
8 of 8
Effect sizes
0.23 to 1.40
Mainly centrebased
Mainly parent-child and parenting
5 of 6
Comprehensive 5 of 6
0.45 to 0.8
0.37 to 1.80

115
110
105
IQ 100
95
<24 mos
>24 mos
90
85
Severely malnourished
Moderately malnourished
Well nourished
Adapted from Winick M, Meyer K, Harris R. Malnutrition and environmental enrichment and early adoption. Science 1975: 190: 1173-75 and Lien N, Meyer K, Winick M. Early malnutrition and “late” adoption: a study of their effects on the development of Korean orphans adopted into American families. Am J Clin Nutr 1977: 30: 1734 –39.

Philippines ECD Program Effects:
Age and duration (marginal PSM)
Armecin, G, JR Behrman, P Duazo, S Ghuman, S Gultiano, EM King, N Lee, and The Office of
Population Studies, University of San Carlos ECD Team, 2006. “Early Childhood Development through
Integrated Programs: Evidence from the Philippines,”, 2006.

• Comprehensive
• Target disadvantaged children
• Begin with younger children
• Sufficient intensity and duration
• Quality
– Training of staff,
– Initiative and exploration in learning environment,
– Partnership with families
– Blend traditional child rearing with evidence-based approaches

• FFE programs aim to increase enrolment, attendance and performance in school.
– School feeding programs (SFP) –subsidy to school attendance and increase in real income for family.
– Take home rations (THR)-can be resold unlike SFP programs.
• Long history of FFE programs, e.g. World Food
Program operating for > 45 years.
*For more details, see Behrman, JR, SW Parker, & PE Todd (BPT), 2011,
“Incentives for Students and Parents”, in P Glewwe, ed. Education Policy in
Developing Countries: What Do We Know, and What Should We Do to
Understand What We Don’t Know? (in preparation)

• Global Food for Education pilot: USDA donated surplus U.S. agricultural commodities for use in school feeding and pre-school nutrition programs in 4000 schools in 38 developing countries between 2001 and 2003.
• World Food Program: operated by United
Nations, 22.6 million beneficiaries in school feeding programs in 2008.

• Large number of studies but few rigorous evaluations of Global Food for Education and
World Food Program interventions.
• BPT focus on recent individual level school feeding projects with rigorous evaluations
(mostly randomized).
• Bangladesh (5 studies), Burkino Faso (2) and
Uganda (2), Jamaica (1) and Peru (1).

• Generally positive impacts of FFE programs on enrollment and attendance.
• More limited results on performance in school including repetition rates and achievement tests.
• Few estimates of real resource costs and relationship to benefits.

• Monetary transfers to poor families conditional on human capital investment, principally regular school attendance of children in beneficiary families.
• Over 30 countries in 5 continents with CCTs, around half with rigorous evaluations.
• Rigorous randomized large-scale evaluation accompanied initial Mexican
PROGRESA/Oportunidades program in late 90s.
– Implementation of many similar programs.
– Impact evaluations implemented in initial program stages.

• Substantial literature on short-run impacts in different contexts.
• Number of reviews of CCTs, most prominent probably 2009 WB PRR (Fiszbein & Schady).
• Behrman, Parker & Todd (2011) focuses mainly on studies carried out since WB 2009.

• Focus on results for short-run educational indicators for 10 countries with CCTs and rigorous evaluations (mostly randomized).
• Significant impacts on enrollment in all contexts with exception of Turkey.
• Impacts range from 2 to 13 percentage points in Latin American countries and 11 to 31 percentage points for countries outside the region (Bangladesh, Pakistan, Cambodia).

• Enrollment impacts higher in countries and contexts with lower levels of schooling.
• Impacts higher for poorer families than for less poor families.
• Evidence on achievement scores and longer-term education indicators “disappointing.”
– Few available studies show no impacts on achievement tests.
• But only 2 based on achievement tests applied in the household.

• Sig impacts on attendance/enrollment for Malawian girls 13-22 y old randomized to conditional transfers but not to unconditional transfers vs. control group
(Baird, McIntosh & Ozler 2010, WB Working Paper).
• Impacts 5-10 percentage points higher for Mexican
Oportunidades families who received enrollment forms vs. those not receiving enrollment forms
(unconditioned)
(DeBrauw and Hoddinott 2010 JDE).
• Morrocan J-PAL study (in process)
• Appears conditionality plays main role in generating educational impacts, but conditionality has costs:
– Individual compliance costs.
– Program enforcement costs.

• Colombia (Bogota): Randomized
(1) traditional
CCTs for attendance, (2) some benefits delayed to beginning of next year & (3) some benefits linked to graduation and reenrollment in next level. Sig impacts on attendance in all 3 but higher impacts on reenrollment under (2) and (3).
(Barrera-
Osorio, Bertrand, Linden and Perez, 2011. American Economic Journal: Applied
Economics).
• Matched control Mexican urban evaluation: enrollment impacts on children 6 to 20, about
3 percentage points, 0.1 grades of schooling.
(Behrman, Gallardo, Parker, Todd and Velez, 2010)

• Gender: Impacts in some cases sig higher for boys (but schooling attainment conditional on some enrollment for girls >= for boys)
(Grant and Behrman 2010 Population Development Review)
• Spillovers on non-beneficiary children a priori could be either direction but in Mexican rural program are positive
(Lalive and Cattaneo, 2009 ReStat and Bobonis and Finan, 2009 ReStat)
• Sig impact on achievement tests in math and English for girls in Malawi
(Baird et al 2010 WB Working Paper).
• Longer-term impacts for initially 9 to 12 y olds after 6 y of benefits in rural Mexico: Boys 1 additional grade, girls 0.8; some evidence of increased work and substitution to non-agricultural work
(Behrman, Parker & Todd,
2011 JHR).

• Popular program, new efforts likely to continue. Evaluation challenging given bundle of goals including both current and long-run poverty alleviation
• Substantial evidence on enrollment impacts.
• Some evidence on nutritional and other health–related impacts.
• Very limited evidence that improve learning and achievement.
• Evidence on longer-term impacts limited.

• Randomized cash grants for performance on matriculation exams in 40 Israeli secondary schools.
Students in treatment schools were 6 to 8 percentage points more likely to pass high school matriculation exam after one year
(Angrist and Lavy, 2009 AER).
• Randomized scholarships for 15% highest scoring girls per district on standardized test in Kenya led to improvements of 0.2 standard deviations on achievement tests. Though students lower in the distribution had little chance of obtaining scholarship, they also showed significant impacts.
(Kremer, Miguel and Thornton, 2009 ReStat)

• Aligning learning incentives (ALI) with randomization among 88 Mexican high schools of alternative monetary incentive payments based on level and improvement in curriculum-based math test scores: students only, teachers only, both students and teachers with own and group components.
After two years if students receive incentives improvement of 0.4 to 0.7 SD, but no significant effect if teachers alone receive incentives.
(Behrman, Parker, Todd & Wolpin, 2011).

• Randomized teacher incentive program in 500 government rural primary schools in Andhra Pradesh,
India. Bonus payments (mean 3% of annual pay) based on average improvement of students' test scores. After two years students in incentive schools had significantly higher test scores by 0.28 and 0.16
SD in math and language, as well on subjects for which no incentives. Incentive schools performed significantly better than other randomly-chosen schools that received additional schooling inputs of a similar value.
(Muralidharan, Karthik and Venkatesh Sundararaman, 2008,
“Teacher Performance Pay: Experimental Evidence from India,: San Diego, CA: U. of
CA and Washington, DC: World Bank )

• Most evaluations of vouchers in developing countries focus on Chile, with universal program introduced in 1981; limited impact on student achievement even in communities where private schools expanded most
(Mizala and
Romaguera, JHR 2000; MacEwan Ed Ec 2001; Hsieh and Urquiola JPubEc 2006)
• Colombia used lotteries to distribute vouchers subsidizing cost of private secondary school (PACES); lottery winners 10 percentage points more likely to finish 8 th grade and with significantly higher achievement tests cores (0.2 SD).
(Angrist, Bettinger, Bloom,
King and Kremer, 2000 AER)

• Apparently substantial short-run impact of ECD programs.
• FFE and CCT incentive programs to students and parents do seem to get kids to go to school more.
• But effects on learning via standardized tests more elusive.
• Perhaps most promising intervention is direct linking of incentives to achievement and other desirable indicators such as graduation.

• Differences in:
– Program objectives
– Initial conditions
– Prices and other policies
• Therefore search for “best practices” needs to be calibrated for such differences

Madagascar
Kenya
India
Mexico
Net Primary School
Enrollments in 2000
68%
65%
79%
97%

1. At minimum such a table must include some of characteristics of key relevance, such as prevalence of worms, misinformation and enrollments. Still may be quite misleading.
2. But perhaps most important lesson of
PROGRESA/Oportunidades and subsequent J-
PAL and other studies is the importance of systematic evaluation of programs on significant scale, NOT that the program should or should not be applied in other contexts that may vary in important dimensions.

2008 Lancet review* of existing studies and new estimates based on five cohort studies (Brazil,
Guatemala, India, Philippines, South Africa) suggests longer-run positive effects.
*Victora, CG, L Adair, C Fall, PC Hallal, R Martorell, L Richter, HS Sachdev, 2008,
“Maternal and Child Undernutrition: Consequences for Adult Health and Human
Capital,” The Lancet 371 (Issue 9609), 340-357.

Summary of Selected Associations between Maternal and Infant Anthropometric
Measures and Adult Outcomes based on Victora et al. (2008)
Adult height
Schooling attainment
3.2 cm for 1 HAZ at age 2
0.7-1.0 cm for 1 cm at birth
0.5 cm for 1 cm of maternal height
0.5 grades for 1 HAZ at age 2 (?)
0.5 grades for 1 WAZ at age 2 (?)
Labor income
0.3 grades for 1 kg at birth
8% for 1 HAZ at age 2(?) males
8-25% for 1 HAZ at age 2 (?) females
Birth weight of offspring 208g for 1 kg for mother at birth
70-80 g for 1 HAZ or 1 WAZ of mother at age 2(?)
*HAZ refers to height-for-age Z scores (i.e., number of standard deviations in the international reference population. WAZ is the weight-for-age Z score.

• Better (atole) and less good (fresco) nutritional supplements randomly available across small number of villages in 1969-77; follow-up surveys in 2002-4 and 2006-7 of children (0-7y) in 1969-77 after about 35 years and their children.

Schooling attainment: Effects found for women only
– Improved by 1.2 grades (0.36 SD)
Reading: For men and women (25-42y), improved
Inter-American Reading scores by 0.28 SD
Cognition: For men and women (25-42y), improved
Raven’s Progressive Matrices scores by 0.24 SD
* Maluccio J, Hoddinott J, Behrman JR, Martorell R, Quisumbing A, and Stein A. 2009 “The impact of improving nutrition during early childhood on education among Guatemalan adults ,” Economic Journal
119 (April), 734 –763.

Exposure to improved nutrition before, but not after
3 years, improved wage rates (income/hour) for men but not women. Exposure from 0–2 years had the greatest impact: Wages increased by US $ 0.67 (CI:
0.16,1.17) or 0.45 SD
*Hoddinott J, Maluccio JA, Behrman JR, Flores R, and Martorell R. Effect of a nutrition intervention during early childhood on economic productivity in Guatemalan adults. Lancet 2008; 371:411-16.

Exposure to improved nutrition in early childhood and income
(in US$) earned per hour; n=602 men; age 25-42 y *
0.70
0.60
0.50
US$/hr
0.40
0.30
0.20
0.10
0.00
P = 0.009
0.67
P = 0.007
0.62
P = 0.406
0.22
0-24 0-36
Window of exposure (months)
36-72
*Hoddinott J, Maluccio JA, Behrman JR, Flores R, and Martorell R. Effect of a nutrition intervention during early childhood on economic productivity in Guatemalan adults. Lancet 2008; 371:411-16.

Exposure for females, but not for males, affected their children 0-12 y. Significant increases in:
•
Birth weight of 116 g (CI: 17 g to 215 g)
•
Height 1.3 cm (CI 0.4 cm to 2.2 cm)
•
Head circumference 0.63 cm (CI 0.37 cm to 0.89 cm)
•
But NOT in weight, BMI, arm circumference, triceps and subscapular skinfold thicknesses
(Estimates robust to alternative treatments of standard errors and attrition.)
*Behrman JR, Calderon MC, Preston S, Hoddinott J, Martorell R and Stein A. 2009. “Nutritional
Supplementation of Girls Influences the Growth of their Children: Prospective Study in Guatemala,”
American Journal of Clinical Nutrition .

Cognitive skills (brain) production functions
Behrman, JR, J Hoddinott, JA Maluccio, E Soler-Hampejsek, EL Behrman, R Martorell, M Ramirez and AD Stein, 2008, “What
Determines Adult Cognitive Skills? Impacts of Pre-School, School-Years and Post-School Experiences in Guatemala”, Washington,
DC: IFPRI Discussion Paper No. 826.
Informative about whether schooling dominant or possibly exclusive determinant of adult cognitive skills , as often
(implicitly) assumed.
Given wage function estimates below, informative of whether cognitive skills a channel through which earlylife nutrition works.
Also informative about other topics, ranging from crosscountry regressions to “Flynn effect” (increasing ability measures over time).
(Estimation IV or 2SLS with good diagnostics & robust to alternative treatment of standard errors/attrition.)

Set 1A Set 1B Set 2 Set 3
Life stage
Representation OLS IV OLS IV OLS IV OLS IV
RCS z scores
E1 Not Stunted at age six 0.270 0.153 0.13 0.51 0.13 0.48
(0.060) (0.284) (0.04) (0.22) (0.04) (0.23)
E2 Schooling attainment 0.22 0.11 0.22 0.13 0.22 0.11
(0.01) (0.03) (0.01) (0.03) (0.01) (0.03)
E3a Skilled job tenure -0.0002 0.02
(0.004) (0.01)
E3b Age at interview in 2002–4 0.19 0.20 0.08 0.11 0.11 0.23 0.11 0.20
(0.09) (0.11) (0.06) (0.07) (0.06) (0.08) (0.06) (0.09)
Age at interview squared

100
-0.33 -0.35 -0.12 -0.20 -0.17 -0.37 -0.17 -0.33
(0.13) (0.17) (0.09) (0.10) (0.09) (0.13) (0.09) (0.13)
Constant -2.70 -2.88 -2.16 -2.01 -2.71 -4.38
-2.71 -3.80
(1.41) (1.92) (0.93) (1.07) (0.93) (1.45) (0.94) (1.56)

NVS z scores
Life stage
Representation
E1 Not Stunted at age six
OLS
0.22
IV
0.366
OLS IV OLS
0.13
IV
0.90
OLS
0.13
IV
0.72
(0.06) (0.263) (0.05) (0.28) (0.05) (0.29)
E2 Schooling attainment 0.15 0.13 0.14 0.17 0.14 0.04
(0.01) (0.03) (0.01) (0.03) (0.01) (0.04)
E3a Skilled job tenure 0.02 0.13
(0.01) (0.02)
E3b Age at interview in 2002–4 0.20 0.288 0.12 0.14 0.15 0.31 0.12 0.10
(0.09) (0.108) (0.08) (0.08) (0.08) (0.11) (0.08) (0.12)
Age at interview squared

100 -0.36 -0.49 -0.21 -0.25 -0.26 -0.49 -0.22 -0.21
(0.14) (0.16) (0.12) (0.12) (0.12) (0.16) (0.12) (0.18)

Most previous empirical investigations of importance of and determinants of adult cognitive skills assume that
(1) cognitive skills produced primarily by schooling
(2) schooling predetermined in statistical sense
But may lead to misleading inferences about impact of schooling & also about importance of pre-schooling and post-schooling experiences on what adults know.

Behrman, Jere R., John Hoddinott, John A. Maluccio and Reynaldo Martorell, 2009, “Brains versus Brawn: Labor Market
Returns to Intellectual and Physical Human Capital in a Poor Developing Country,” Philadelphia, PA: University of
Pennsylvania, mimeo.
Vast literature assessing “return” to intellectual human capital (typically schooling) in labor market
(Psacharopoulos & Patrinos 2004)
Some of it incorporates physical human capital (shortterm measures or height), often for developing countries
(Strauss 1986; Thomas & Strauss 1997
)
Other studies incorporate more proximate measures, particularly for intellectual human capital
(Boissiere, Knight
& Sabot 1985; Murnane, Willet & Levy 1995; Alderman et al. 1996; Glewwe et al.
1996)
Here endogenize proximate measures of two HC types

Table 4: Ln hourly wage rate production functions (N=962)
OLS IV OLS IV
(I
1
) Completed grades of schooling
(I
2
) Adult RCS Z score
(H
1
) Ln adult height
(H
2
) Ln adult fat-free body mass
0.084
2.117
0.114
-1.765
0.228
1.079
0.404
-0.553
Male 0.207
0.484
0.102
0.524
Age
R
2
/Centered R
2
0.014
0.19
0.016
0.16
0.012
0.16
0.018
0.11
Kleibergen-Paap test 11.71
8.34
Hansen J test
p-value
8.42
[0.39]
7.96
[0.44]
Notes: Instrumental variables GMM estimates using instruments and second stage controls indicated in Table 3. Standard errors calculated allowing for clustering at the birth year-village cohort level (64 clusters)s.
Significance 0.01
0.05

Benefits
•
Recognize benefits over life cycle and across generations, conditional on survival probabilities, but avoid double counting
•
Estimate benefits, not associations, given behavioral choices
•
Monetary value on each benefit (challenge, e.g. survival)
•
Obtain present discounted value of benefits by discounting
Costs
•
Include private and public resource costs, including distortion costs of raising public resources but not transfers
•
Obtain present discounted value of costs by discounting
Behrman JR, Alderman H and Hoddinott J. “Hunger and Malnutrition” in ed. Bjørn Lomborg, Global
Crises, Global Solutions , Cambridge, UK: Cambridge University Press, 2004, 363-420.

Multiple impacts over life cycle, evaluation (e.g., survival), no double counting (e.g., schooling), discount rates (e.g., table)
Estimates of Present Discounted Values of Seven Major Benefits of Moving One
Infant Out of Low Birth Weight Status in Low-Income Country (U.S. Dollars)
Annual discount rate
Benefit 3% 5% 10%
1. Reduced infant mortality
6. Reduced costs of chronic diseases
7. Intergenerational benefits
Total
Share of total at 5 percent discount rate (percent)
$95 $99 $89
2. Reduced neonatal care
$42 $42 $42
3. Reduced costs of infant and child illness $36 $35 $34
4. Productivity gain from reduced stunting $152 $85 $25
5. Productivity gain from increased cognitive ability $367 $205 $60
$49
$92
$15
$35
$1
$6
$832 $510 $257
163 100 50
Alderman H and Behrman JR. Reducing the Incidence of Low Birth Weight In Low-Income Countries has Substantial Economic Benefits, World Bank Research Observer 21(1): 25 –48, 2006.

Combining Benefits and Costs (Copenhagen Consensus )
Opportunities and targeted populations Benefits/
Costs
1. Reducing LBW for pregnancies with high probabilities
LBW (particularly in South Asia)
1a. Treatments for women with asymptomatic bacterial infections 0.6-4.9
1b. Treatment for women with presumptive STD
1c. Drugs for pregnant women with poor obstretic history
2. Improving infant and child nutrition in populations with high prevalence of child malnutrition (fairly widespread in poor populations)
2a. Breastfeeding promotion in hospitals in which norm has been promotion of use of infant formula
2b. Integrated child care programs
2c. Intensive pre-school program with considerable nutrition for poor families
3. Reducing micro nutrient deficiencies in populations in which they are prevalent
3a. Iodine (per woman of child bearing age)
3b Vitamin A (pre child under six years)
3c. Iron (per capita)
3d Iron (pregnant women) c
1.3-10.7
4.1-35.2
Size of Targeted Populations
12 million LBW births per year
162 million stunted children
4.8-7.4 0-5 years old
9.4-16.2
1.4-2.9
15-520 2 billion people with iodine deficiencies
4.3-43 128 million pre-school children
176-200 3.5 billion people, incl. est. 67 million
6.1-14 pregnant women
Behrman JR, Alderman H and Hoddinott J. “Hunger and Malnutrition” in ed. Bjørn Lomborg, Global
Crises, Global Solutions , Cambridge, UK: Cambridge University Press, 2004, 363-420.

• Benefit/cost of eliminating preschool enrollment gaps among income quintiles for
73 developing countries with 2.69 billion people
(Engle et al. 2011 Lancet)
• For every percentage point increase in preschool enrollment, the schooling gap for
15-19 year olds declines 0.026 grades [5% -
95% CI = -0.14, -0.38], controlling for countries’ GDP and inequality (GINI index)
(robust to using country fixed-effects).

• Economic benefit PDV of added wage productivity for real discount rates of 3% and 6% under assumption that earnings are 0 for first 12 y after preschool & equal to yearly average earnings incremented by average rates of return to schooling (8.3% for urban areas, 7.5% for rural areas) for following 45 y.
• With 3% discount rate, benefits from $10.6 billion with an increase of all children in each country to 25% enrollment for preschool to $33.7 billion with increase to 50% preschool enrollment. With a 6% discount rate benefits are $4.7 billion (for 25%) to $14.9 billion (for
50%).
• Imply benefit/cost ratios respectively from 6.4 to 17.6

• CCTs
• Vouchers

3% 3.60
5.36
7.11
5% 1.70
2.77
3.84
Behrman, Jere R., Susan W. Parker, and Petra E. Todd, 2011, “Do Conditional Cash
Transfers for Schooling Generate Lasting Benefits? A Five-Year Follow-Up of
Oportunidades Participants, Journal of Human Resources 46:1 (Winter), 93-122.

Structural model to identify impacts from differences in schooling and work choices made and wage returns received by individuals differentially exposed to program and find: vouchers induced higher attendance at private subsidized schools, higher schooling attainment, higher wages and higher labor force participation
(all more for non-poor). Returns to both public and private schools increased.
(Bravo, David, Sankar
Mukhopadhyay and Petra E. Todd, 2010, “Effects of a Universal School Voucher
System on Educational and Labor Market Outcomes: Evidence from Chile,”
Quantitative Economics )

• Colombian vouchers allocated by lottery in cities with excess demand (PACES): no significant impact on enrollment or drop-out, 15 percent more likely to attend private school, 0.12-0.16 more grades of schooling (primarily due to lower repetition rates), ~10 percentage points more likely to complete 8 th grade, 0.2 standard deviations higher on standardized tests, worked
1.2 hours per week less, less likely to be either married or cohabiting as teenagers, 15-20% higher secondary school completion rates; benefit-cost ratios of 3.8 with a discount rate of
3%, 2.7 with a discount rate of 5%, and 1.4 with a discount rate of 10%
(Angrist et al. 2002, Angrist, Bettinger &
Kremer 2004, Knowles & Behrman 2005).

• India has 22.6% low birth weights vs. 8.9% for
US. Closing gap by increased all birth weights
< median in India would increase earnings by estimated 9.2%. But closing gaps for all countries for which relevant data would reduce world income inequality by < 1%.
(Behrman, Jere R. and Mark R. Rosenzweig, 2004, “Returns to Birthweight,” Review
of Economics and Statistics 86:2 (May), 586-601.)

Targeted Schooling and Earnings Distribution
– Illustration with 2004 Chilean SPS
(Behrman, Jere R., 2011, “How Much Might Human Capital Policies Affect
Earnings Inequalities and Poverty?” Estudios de Economia (June))
• Given qualified, but definitely positive assessment of a number of HR-related interventions, of interest to ask what would be impacts on income distribution were there some targeted increases in human capital
• Simulate impacts of hypothetical changes in schooling attainment; focus on schooling attainment because it is most-emphasized among possible human capital investments -- indeed many studies empirically seem to equate schooling attainment with human capital

Basic description of inequality and “poverty” in the 2004 Chilean
SPS:

• HR targeted towards lower part of distributions often thought to be major means to reduce inequality and poverty, increased knowledge of policy effects.
• B/C estimates for range of HR interventions, though with caveats (e.g., due to sensitivity to discount rate, value of life), suggest opportunities for range of HRs
(not just schooling)
• Will targeted HR substantially reduce inequality?
Simulations of schooling suggest (a) significant effects, (b) relevance of targeting, (c) range of possible effects on poverty – but (d) magnitudes of effects on inequality not all that great

•
•
•

Japan
EC
United Kingdom
Germany
France
Netherlands
Denmark
UNDP
Italy
Canada
Ireland
Belgium
Luxembourg
Portugal
Norway
New Zealand
UNICEF
Australia
Finland
Spain
0 5 10 15
Aid to ECCE as
% of aid to primary education
Almost all donors allocate to pre-primary less than 10% of what they give to primary
Bilateral donors give priority to center-based program for children from age 3.
From EFA Report, UNESCO 2007

Criteria for program evaluations in Engle et al 2007:
• Conducted since 1990
• Randomized controlled trial or matched groups (i.e. marginal PSM)
• Intervention before 6 years
• Effectiveness or program evaluation (scale)
• Child development assessed
• Disadvantaged children
• Developing country
20 studies located met these criteria

• Generally significant impacts of FFEs on school enrollment:
– effects range from 6 to 26 percentage points.
• Also significant positive effects on attendance in 5 of the 10 studies.
– effects range from 2 to 15 percentage points.
– But one study (Kazianga et al., 2009 for Burkina
Faso) finds negative effects on attendance, concentrated on households with low availability of child labor in the household.

• Mixed effects on progression and repetition indicators and test performance.
– Bangladesh THR reduced dropout in 9% but
Ugandan THR did not.
– Philippines SFP no effect on dropout rate, but significant reductions of Ugandan SFP in reducing grade repetition for boys.
– Mixed pattern on achievement tests, at least 2 programs show some significantly negative effects on achievement.

• Large-scale non-linear incentive scheme for teachers in 950 schools in Pernambuco Brazil in 2008. Find differential response across subjects (0.3 SD for Portuguese > response for math) and grades (greater for 8th than 12 th grade and non-linear contracts induce both positive and negative effects. Ferraz, Claudio and Barbara Bruns, 2011,” Incentives to Teach:
The Effects of Non-Linear Incentive Contracts in Brazilian Schools”, Rio de Janeiro, PUC and
Washington, World Bank.

• Increasing body of evidence regarding substantial impacts of life cycle of ECD – and some regarding benefits/costs
• What Nobel Laureate J. Heckman (2006) said for ECD in US also probably true for developing countries: “It is a rare public policy initiative that promotes fairness and social justice and at the same time promotes productivity in the economy and in society at large.
Early interventions targeted to disadvantaged children have much higher returns than later investments.
”

Supplementation period
Birth
Year
1962 1969 1974 1977
Too old for complete exposure for 0 –36 m
Complete exposure for
0-36 m
Too young for complete exposure for
0-36 m

Average outcome for those exposed to atole 0 –36 m completely
–
Average outcome for those exposed to fresco 0 –36 m completely
Average outcome for those NOT exposed to atole
0 –36 m completely
–
Average outcome for those NOT exposed to fresco
0 –36 m completely

Impact estimates: right-side behavioral variables and unobserved determinants
• For example, identifying impact of stunting at 36 m on subsequent outcomes if stunting proxies in part for unobserved family background, etc.
• Critical data attributes and estimation options
 Baseline and longitudinal follow-up (ideally for all impacts, many years)
 Representative of relevant population (e.g., not selected by clinic use)
 Sufficient sample size for power
 Treatment and control groups:
 Random assignment and difference-in-difference estimates
 Propensity score matching
 Regression discontinuity
 Change in right-side behavioral variable (instruments for IV estimates)

Semilog earnings & wage rate functions based on the 2004 SPS

• Schooling inequality: Generally down (but sim1), about twice as much if targeted to low schooling, most if bring all up to 6 grades (norm Gini -0.08)
• Schooling “poverty” headcounts: Similar targeting point but with nuance (Sim 7); relatively effective compared to inequality (norm HC -0.-9 to -0.33)
• Earnings inequality: Targeting point reversed (1.5 to 3 times as effective) with drop in norm Gini -
0.03– but how to do?
• Earnings poverty headcounts: first 4 simulations have same impact on norm Gini but very different on norm HC (-0.03 for sim 4 to -0.08 for sim 1)

• Wage rate inequality: Similar to earnings inequality but norm Gini drops a little more
• Wage rate poverty headcount: Relatively effective once again, but preferred simulation depends on what cutoff