The Entry of NGO Schools and Girls’ Educational Outcomes in Bangladesh

Pataporn Sukontamarn

London School of Economics

November 29, 2004

Outline

Research question and motivation Theoretical framework Background and data Empirical analysis Conclusions

Research Question

How does the entry of NGO schools affect the educational outcomes of girls in Bangladesh?

Does the entry of NGO schools increase girls’ enrollment relative to boys?

Characteristics of NGO and state schools that affect girls’ enrollment

How does attending an NGO school affect test scores?

Motivation

1. Rapid increase in girls’ enrollment rates in recent years

=> Gender equity in primary enrollment

2. Heavy involvement of NGOs in the provision of non-formal primary schools

1.4 million children Aim: poorest children (non-enrolled or

dropout) Many different characteristics

Theoretical framework

Show how NGO schools might affect girls’ enrollment as compared to boys’

Schooling decision in cost-benefit framework

Main assumption: disutility of sending a child to school, which differs by gender and by school type

Max B(xs) – psjk + U(Y-ckj-xs) (1) {xs}

B(•) = benefits of educationU(•) = utility from consumptions = male (m), female (f)k = no school (0), gov. sch. (g), NGO sch. (n)j = village jxs = expenditure on education of child spsjk = disutility of sending child s to school k in village jY = incomeckj = opportunity cost of time of going to school k

Assumptions:B(•) increasing & concave in xs

U(•) increasing & concave in Ypfgj > pmgj = pmnj = pfnj

cgj > cnj

Let xj*(Y) be solution to (1) and V(Y) be maximum value function

Enrol a child if:

V(Y) - U(Y) > 0

Results:

• Enrollment ↓ in pjk and cj and in Y

Let Yjk* be threshold level Y above which a household will enrol child j in school k, thenYmnj* = Yfnj* < Ymgj* < Yfgj*

• NGOs contribute to increase in enrollment of boys and girls, with stronger effects for girls

Assume NGOs target households Y < Ytg

Suppose Ymn* = Yfn* < Ymg* < Ytg < Yfg*

• Gender gap disappears for the poor, but exists for the rich

Background

BRAC : largest NGO (1.2 million out of 1.4) BRAC schools:

One-room school built inside village 33 children & 1 teacherOver 90% of teachers: femaleClass time: decided by parents & teacher 3-year cycle covering 3 years of gov. school

curriculum

Data Education Watch 1998 3 survey instruments: (1.) Household Survey Questionnaire

(42,584 households, 31,092 children)(2.) Assessment of Basic Competencies

(ABC) (3360 children)(3.) School Observation Checklist

(885 schools)

Empirical analysis

(1.) Entry of NGO schools and girls’

enrollment

(2.) School characteristics that affect girls’

enrollment

(3.) Test scores

(1) Entry of NGO schools and girls’ enrollment: an overview

TABLE 2 PERCENTAGE OF CHILDREN WITH NO SCHOOLING

BY COHORT AND VILLAGE OF RESIDENCE

NGO school availability in village of residence At least 1 NGO school No NGO school Aged 17-20 Aged 11-14 % Change Aged 17-20 Aged 11-14 % Change

(1) (2) (3) (4) (5) (6)

Girls: .273 .126 -53.8 .244 .149 -38.9 (.44) (.33) (.43) (.36) Boys: .234 .180 -23.1 .207 .165 -20.3 (.43) (.39) (.41) (.37)

Standard deviations are in parentheses.

(1.1) Exposure to NGO schools and enrollment (11-20 years old)

Pr(Sij = 1) = α0 + α1EXPij + α2Girl*EXPij + α3Cij + α4Girl*Cij + α5Girl + α6Vj +error

term

where,Sij = 1 if individual i in village j was enrolled

at time of survey, and 0 otherwiseEXPij = 1 if individual i in village j was exposed to an NGO school in the villageGirl = the dummy variable for being a girlCij = a vector of child and family characteristicsVj = village fixed effects

TABLE 3 ENTRY OF NGO SCHOOLS AND GIRLS’ ENROLLMENT

11-20 years old

Dep var: Ever enrolled = 1

(1) (2) (3) (4)

Exposed to NGO school .178 .004 .218 .087 (1.9)* (.03) (1.9)* (.70) Girl*Exposed to NGO school .385 .302 (2.9)*** (2.5)** At least 1 NGO school .034 -.043 (.39) (.48) Girl*At least 1 NGO school .165 (1.9)* At least 1 gov. school .181 .100 (1.5) (.75) Girl*At least 1 gov. school .159 (1.1) Control for village No No Yes Yes characteristics Stratum effects No No Yes Yes Village effects Yes Yes No No Number of observations 43738 43738 43738 43738 Pseudo R2 .255 .258 .216 .219

(1.2) Involvement of NGO schools and enrollment (6-10 years old)

• For each child, define: ‘the involvement of non-formal schools in a village’

number 6-10 yrs old in NGO schoolsnumber 6-10 yrs old enrolled in school

excluding the child from the sample

‘ the involvement of government schools in a village’ is similarly defined

Pr(Sij=1) = α0 + α1Nij + α2Girl*Nij + α3Gij + α4Girl*Gij +α5Cij + α6Girl*Cij

+ α7Girl+ α8 Aj + α9Vcj + error term

where,

Girl = dummy variable for being a girl

Girl*Nj shows the difference in the effects that the involvement of NGO schools has on the enrollment of girls as compared to boys

TABLE 4 NGO SCHOOLS AND GIRLS’ ENROLLMENT

6-10 years old

Dep var: Currently enrolled = 1

(1) (2) (3) (4)

At least 1 NGO school .013 -.114 (.15) (1.2) Girl*At least 1 NGO school .220 (2.6)*** At least 1 gov. school .107 .089 (.71) (.54) Girl*At least 1 gov. school .045 (.36) Fraction: NGO schools -.656 -.917 (1.7)* (2.2)** Girl*Fraction NGO .543 (1.8)* Fraction: Gov. schools .075 -.006 (.47) (.04) Girl*Fraction Gov. .168 (1.3) Control for child and Yes Yes Yes Yes family characteristics Number of observations 31033 31033 31033 31033 Pseudo R2 .143 .146 .143 .146

- Rural versus Urban

- BRAC target versus non-target

Definition:

BRAC target: households with less than 0.5 acre of land and at least one

person engaged in manual labor for at least 100 days per year

(1.3) Extensions:

TABLE 5 SCHOOL ENROLLMENT: BOYS AND GIRLS AGED 6 TO 10

Dependent variable: Currently enrolled in school = 1

Rural versus urban Target versus non-target

Rural Urban Target Non-target (1) (2) (3) (4)

Fraction: NGO schools -1.02 -1.88 -.737 -1.39 (1.9)* (2.5)** (1.1) (3.4) Girl*Fraction NGO .796 -.086 .872 .648 (2.2)** (.22) (2.0)** (1.6) Fraction: Gov. schools -.198 -.025 -.029 -.364 (1.1) (.06) (.14) (1.8) Girl*Fraction Gov. .116 .015 .198 -.026 (.85) (.06) (1.1) (.14) Girl -.591 -.305 -.684 -.216 (1.8)* (.60) (1.6) (.66) Control for child and Yes Yes Yes Yes family characteristics Stratum effects Yes Yes Yes Yes Number of obs. 24533 6500 10331 14202 Pseudo R2 .149 .186 .107 .142

(2) School characteristics that affect girls’ enrollment

Pr(Sij = 1) = α0 + α1Ncj + α2Girl*Ncj + α3Gcj + α4Girl*Gcj + α5Cij + α6Girl*Cij + α7Girl + α8Aj + α9Vcj + error termwhere,Ncj = a vector of aggregate village-level characteristics of NGO schools in village jGcj = a vector of aggregate village-level characteristics of government schools in village j

TABLE 7 SCHOOL CHARACTERISTICS AND ENROLLMENT:

BOYS AND GIRLS AGED 6 TO 10

Dependent variable: Currently enrolled in school = 1

All Bangladesh Rural areas

Pooled Interacted Pooled Interacted Level *Girl Level *Girl

(1) (2) (3) (4) (5) (6)

Class size (NGO) .012 .004 .016 .029 .005 .056

(.67) (.19) (.78) (1.3) (.22) (1.5)

Class size (Gov.) -.008 -.007 -.001 -.014 -.013 -.002

(2.5)** (2.2)** (.18) (2.9)*** (2.4)** (.37)

% teachers present (NGO) -.577 -.304 -.811 -1.02 .107 -3.06

(.22) (.11) (0.38) (.28) (.03) (1.20)

% teachers present (Gov.) -.254 -.096 -.497 -.381 -.032 -1.06

(.48) (.17) (1.1) (.54) (.05) (1.9)*

% female teachers (NGO) 1.41 .927 1.04 1.29 .924 .681

(4.3)*** (2.7)*** (2.3)** (2.4)** (1.7)* (1.3)

% female teachers (Gov.) .165 .007 .276 .358 0.63 -.794

(.30) (.01) (.67) (.58) (1.1) (1.6)

Teachers' education (NGO) -.026 .002 -.056 .060 .071 -.008

(.31) (.02) (.89) (.67) (.78) (.11)

Teachers' education (Gov.) .056 .003 .133 .068 .009 .159

(.39) (.02) (1.2) (.40) (.05) (1.3)

Teachers' experience (NGO) -.013 -.027 .019 -.012 -.018 -.003

(.44) (.67) (.47) (.32) (.36) (.06)

Teachers' experience (Gov.) .034 .006 .065 .047 .028 .048

(1.3) (.23) (2.7)*** (1.6) (.98) (2.2)**

% of schools with PTAs (NGO) .367 .364 -.005 .591 .530 .163

(1.8)* (1.7)* (.03) (2.0)** (1.7)* (.95)

% of schools with PTAs (Gov.) -.047 -.025 -.078 -.010 .011 -.087

(.28) (.15) (.50) (.04) (.04) (.51)

% of schools with SMCs (NGO) .316 .474 -.331 .063 .140 -.176

(1.4) (2.0)** (1.7)* (.22) (.51) (.87)

% of schools with SMCs (Gov.) 1.84 2.72 -1.97 6.72 7.39 -.985

(1.5) (1.9)* (1.5) (2.0)** (2.3)** (.27) Number of observations 8611 8611 7045 7045 Pseudo R2 .158 .168 .159 .172

(4) Test scores Yij = α0 + α1Tij + α2Girl*Tij + α3Cij + α4Girl*Cij +

α5Vj + error term

where,

Yij represents

(i) whether a child passed the ABC test or not, and

(ii) test scores of life-skills, reading, writing, and

numeracy sections

Tij = dummy variable for the type of school that the

child was attending at the time of survey

TABLE 8 TEST SCORES: BOYS AND GIRLS AGED 11 TO 12

ABC Life skills Reading Writing Numeracy

Level *Girl Level *Girl Level *Girl Level *Girl Level *Girl

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Currently attending .902 2.02 .708 .498 .838 .053 2.35 .468 .384 .105 NGO school (1.6) (2.45)* (2.6)** (1.4) (4.6)** (.20) (5.6)** (.79) (2.3)* (.45) Currently attending -.046 1.506 -.132 .269 .361 -.030 0.89 .015 .212 -.068 government school (.10) (1.9) (.56) (.88) (2.6)* (.16) (3.1)** (.04) (1.4) (.30) Indicator: Girl -1.48 -.850 .604 .383 -.943 (1.1) (1.5) (1.9) (.73) (4.0)** Control for other family charac. Yes Yes Yes Yes Yes and other school types Village effects Yes Yes Yes Yes Yes Number of observations 3061 3324 3324 3324 3324 Pseudo R2/ Adjusted R2 .339 .363 .518 .577 .295

Conclusions The entry of NGO schools significantly

increases girls’ enrollment relative to boys.

It is mainly in the rural areas that NGO schools appear to have strong effects.

The effects of NGO schools are stronger for BRAC target households compared to non-target households.

The main characteristics of NGO schools that appear to encourage girls’ enrollment: high percentage of female teachers and having PTAs

Attending NGO schools shows significant effects on ABC test scores.