Learning and Educational Attainment
in Punjab, Pakistan:
Thoughts for Debate and Discussion
Growth Commission on Education
Jishnu Das (CPR & World Bank)
Tahir Andrabi (Pomona College)
Asim Khwaja and Tristan Zajonc (Harvard U.)
Tara Vishwanath (World Bank)
South Asian Education in the New
Millennium
„ Remarkable growth of private schools in
South Asia during the last 10 years
A map of India
Implications
„ This changes
Evaluations of government policy towards
schooling (worry about G.E. effects)
„ Construction of research questions (what is
the right intervention?)
„ The policy framework for debate (a modified
role of the government)
„
Yet…
„ We know very very
little
„ Organization of the “village” education market
„
Village is never taken as sampling unit for
facility surveys
„ Parental choice and optimization
„
Little data linking household decisions and
school-level inputs
„ The market for teachers
„
Little information on supply elasticity
Data framing the presentation
„ First year of 4-year longitudinal study in 112 villages
„
„
„
„
„
Randomly selected from villages with a private
school in 3 districts of Punjab
All schools in these villages: 812 public and
private schools
5000 teachers
12000 children followed and tested annually for 4
years
2000 households with 900 children linked to testscore information
Why Punjab?
„ The heartland of private schooling
„ Strong, pro-active lending program from The
World Bank
Survey baseline initiated with WB program
„ Follows outcomes every year of the program
„ Builds in randomized evaluations
„
„
„
Information
School-Management Councils
This Presentation
„ Insights from the Learning and Educational Attainment Project in
Punjab, Pakistan
„ Basic descriptive facts about schools, teachers, households and
learning in villages from 1st year of survey
„ Relevant highlights about
„ Schools
„ Teachers
„ Households
„ Learning Outcomes
„ Set the stage for academic research
„ …and a discussion on the policy framework
A Bit About Pakistan
„ Four main provinces: Punjab, North-West
Frontier Province (NWFP), Balochistan and
Sindh account for 97% of population
„
Punjab is the largest (55% of population)
„ Azad Jammu Kashmir (AJK), FATA,
Islamabad and the Northern Areas are the
other 4
A Bit About Pakistan
„ Traditionally poor educational numbers
„ But enrollment is looking up
„
Net Enrollment rates increased from 51 to 61
percent between 2001 and 2005
„ Like other countries, discourse based on
household surveys and school censuses
Little information about learning
„ No links between households and schools
„
The “new” educational landscape
„ Rapid growth of private schools
„ Questions much like in the US
(vouchers/regulation/failing government
schools),
„
except for religious schooling
„
which plays a much smaller role: 1% of all
enrollment in 2001 (9% in the US)
Private Schools
„ Huge growth in number of schools from 1995
„ 3800 in 1983, 32,000 in 2001 and 47,000 in 2005
„ Reflected in large increases in enrollment
„ Share of private schooling increased from 29.6% to
36.9% (Punjab), 19.9% to 25.2% (NWFP) and 30.6%
to 33.5% (Sindh)
„ Median private school was $1/month in 2001 and
$1.5/month in 2005
„
Highest enrollment growth in rural areas and lowmiddle income households ( 1991-2001)
Private Schools…the but…
„ Primarily a Punjab and NWFP phenomenon
„
Not rural Sindh or Balochistan
„ Primarily a richer village phenomenon
„
Although this seems to be changing slowly
„ Government schools are everywhere
„
And the only option in most poorer villages
Distribution of Private Schools
Population per Private School Across Village Socio-economic
Categories
Persons per private
school
20000
15000
17620
2001 (Private)
2005 (Private)
13169
2005 (Public)
10119
10000
6033
3673
5000
629
712
2486
832
0
Lowest SES
Middle SES
Highest SES
Village Type
Private schools are overwhelmingly located in high SES villages while
there is equal access to government schools in all types of villages
LEAPS data adds the within village
perspective to this overall picture
and looks at all inputs into the
education production function
Here is what a village looks like now
50.44
GPS Longitude East (Decimal)
50.445
50.45
50.455
50.46
50.465
One Village inAttock
g
13
6
21
7
4
g 2
3
50.69
1
5
25 mins walk
50.695
50.7
50.705
11
8
50.71
50.715
GPS Latitude North (Decimal)
Govt. Boys’ School
Govt. Girls’ School
Private School
These location patterns are generic
sc
ho
ol
s
sc
ho
ol
s
e
e
6
or
m
or
m
or
or
e
m
2
m
1
% Private Schools
35
sc
ho
ol
or
e
Sc
ho
ol
nl
y
O
sc
ho
ol
s
% Private and School Wealth Index
-.5
0
.5
1
School Locations and Number of Nearby Schools
School Wealth Index
Private schools are overwhelmingly located in schooling clusters
where students are richer
Tremendous Variation Across Schools
„ Infrastructure and student-teacher ratios
Figure 2.4: Private schools report lower student-teacher ratios and better basic facilities…and there
are some government schools where student-teacher ratios exceed 50 and basic facilities are
completely absent
Student/Teacher Ratio in Schools
Basic Facilities By Type
0
0
.02
kdensity
.04
kdensity basic_facilities
.1
.2
.3
.06
.4
By School Type
0
100
200
300
400
x
Government Schools
500
-2
0
2
4
6
x
Private Schools
Government Schools
Private Schools
8
Tremendous Variation Across Schools
(2)
0
Rs. per year per student
.0005
.001
.0015
The Distribution of Teaching Costs
0
1000
2000
3000
4000
5000
x
Government Schools
Private Schools
The median cost of educating a child in the government sector, at
Rs.2000 per year, is twice that in the private sector, at Rs.1000
So what drives cost-differences?
Teacher Compensation
6,000
6178
Salary in Rs.
2,000
4,000
5299
1619
0
1231
Unadjusted
Adjusted
Private
Unadjusted
Adjusted
Public
Private School Teachers are paid 20% of their government
counterparts
Hiring and compensation
„ Government Sector
Hires more educated and better trained
teachers
„ Ensures equitable distribution across all
villages
„
„
Setting wages to ensure access in remote areas?
Has no link to performance
„ Pays a uniform wage (no discrimination)
„
Compensation and the labor market
Public/Private Wage Premiums and Penalties
Public School Training
Premium
Public Female Penalty
113.8
2.0
Public Local Penalty
-5.8
1
Private School Training
Premium
3.0
-28.1
Private Female Penalty
-20.5
-50.0
Private Local Penatly
0.0
50.0
100.0
150.0
Percent Premium/Penalty
The private sector exploits local market conditions: women and local
teachers are paid (a lot) less
Private Sector Location Decisions
.05
.1
Probability
.15
.2
Fi gu re II. P rob a b il i ty of P ri va te Sch oo l w/ E xp osu re to Go ve rnm e nt S cho o l s
0
5
10
E xpos ure in Y ears
15
Girls P rim ary S c hool E xpos ure
Girls S ec ondary Sc hool Expos ure
B oys P rim ary S c hool E xpos ure
B oys S ec ondary Sc hool Expos ure
20
As a consequence, private sector location decisions respond to the
availability of an educated female workforce
How do households make decisions?
„ Information
„ Household Preferences
Information: Parents know how well
their kids are doing
Mother's Perception of Intelligence and Child's Test Score
-.6
-.4
Scaled Test Score
-.2
.2
0
.4
By Literacy of Mother
Poor
Average
Above Average
Poor
Illiterate
Average
Above Average
Literate
English
Urdu
Math
Information: Parents know how good
schools are
Parental Assessment and Actual Performance of Schools in English
By Literacy
Fathers
-.4
-.2
0 Scaled Test .2
English
Score
.4
Mothers
Cannot Read
Poor
Can Read
Average
Above average
Cannot Read
Poor
Can Read
Average
Above average
Preferences: Distance matters
0
.4
.5
.6
.7
% Enrolled
Percent of All Schools
5
10
.8
.9
15
Distance and Enrollment
0
.5
1
Distance to Closest School
Frequency of Distance to School
Boys
1.5
Girls
For girls, every 500m distance from school is
associated with a 15%-point drop in enrollment
Preferences: Parents choose winners
0
Mean Education Expenditure
100
200
300
400
Education Expenditure by Intelligence and School Type
Private
Public
Very Poor
Poor
Above Average
Highly Above Average
Average
So: What do they learn
„ Across all schools, children perform far below
curricular standards by the end of Grade III
Majority can do single-digit subtraction
„ Recognize English alphabets and simple
picture matching
„ Read simple words in Urdu, but only 31
percent can write a sentence with the word
“school”
„
Learning Outcomes: The Facts
Table 1.1: What do Children Know in Mathematics
Subject
The Question
Percentage
who answered
correctly
Corresponding
Grade for
Curriculum
Mathematics
Mathematics
Mathematics
Mathematics
Mathematics
Mathematics
4+6
89
K&I
36 + 61
86
K&I
8–3
65
K&I
5x4
59
II
238 - 129
32
II
Read and Write the time (Clock
24
II
shows 3:40)
Mathematics
19
III
384 ÷ 6
Mathematics
4 x 32
50
III
Mathematics
Fractions: ½ + 3/2
19
III
Mathematics
Read a diagram of a scale to answer
12
III
which part is heavier
Mathematics
Fractions: 7/5 – ¾
1
IV
Notes: Questions are from the Pakistan LEAPS exam. Columns 3, 4, and 5 report the
percentage of children who answered the question correctly in all schools, government
schools, and private schools.
Children are performing significantly below curricular standards in
Mathematics
Learning Outcomes: The Facts
Private−Government School Gap
149
150
143
101
100
92
76
73
50
0
Math
Urdu
Unadjusted Gap
English
Adjusted Gap
The main learning differences are across public and private schools
Learning Outcomes: The Facts
Rich−Poor Gap
80
Father Literate−Illiterate Gap
80
73
66
60
40
60
49
45
40
40
20
15
12
7
0
34
20
11
9
7
0
Private−Government School Gap
149
150
143
−20
101
100
92
76
73
Math
Urdu
English
−20
Math
Mother Literate−Illiterate Gap
50
0
Math
Urdu
Unadjusted Gap
English
80
Urdu
English
Female−Male Gap
80
Adjusted Gap
60
57
60
47
40
40
35
27
20
4
5
7
0
17
Math
Urdu
English
Unadjusted Gap
18
4
0
−20
−20
26
20
−12
Math
Urdu
English
Adjusted Gap
Differences across public-private schools are 8 to 15 times the difference
across socio-economic categories
Learning Outcomes: The Facts
English
Urdu
Math
Class 3 Private
600
597
596
Private−Government School Gap
149
150
143
101
100
92
76
73
50
0
Math
Urdu
Unadjusted Gap
English
Adjusted Gap
Knowledge Score
568
550
Public
500
507
492
503
455
450
470
3
4
5
3
478
4
5
3
4
5
Class
And its going to take government school children 1.5-2.5 years to catch
up to private school learning outcomes in Grade III
Learning Outcomes: The Facts
0
Cost in Rupees for Every Percent Correct
100
200
300
400
The Cost of Schooling
Public
Private
English
Urdu
Math
The cost per percentage correct in a test is 2-3 times higher in public
schools
…and the worst private schools are
much better than the worst govt.
schools
English Test Scores for Children from Different Families
Average Score in Sample of
12000 children
P erform ance
P erform ance
H igh
H igh
English Test Scores in Public and Private Schools
Average Score in Sample of
12000 children
Public
Private
Schools
Schools
Top Wealth Decile/
Literate Parents
Bottom Wealth
Decile/Illiterate
Parents
Worst Schools 5th
Percentile)
Low
Low
Best Schools (95th
Percentile)
When government schools fail, they
fail completely
SocioEconomic
Status
Literacy
A Proposal for Reform
„ Government sector should complement
private sector
„ Three ways
Provide information for competition (learning
outcomes in every school)
„ Reform government hiring and compensation
of teachers
„ Experiment with vouchers
„
Teacher and Child Test Scores
.5
Private and Public Schools
Child English Test Scores
.3
.4
The difference in test-scores between
children in private and public schools
with secondary educated teachers
.2
The English Test score of a
teacher with a bachelors
.7
.75
.8
.85
Teacher English Test Scores
Private Teachers
The difference in test-scores
between children with a secondary
educated teacher and a teacher with
a bachelor’s degree in the public
school
.9
.95
Public Teachers
The English Test score of a
teacher with secondary education
The figure is based on a non-parametric
plot of child English test scores against
teacher’s English test scores in the public
and private sectors.
Government Teacher Reform Required? (2)
This is the gap in child test-scores between
children in public and private schools, both with
teachers who have a secondary education
This is the gap in child test-scores between
children with teachers who have a secondaryeducation versus a bachelor’s degree
This is the average salary of a teacher in the
public sector
This is the average salary of a teacher in the
private sector
Government Teacher Reform
Required?
„ Reforming hiring and compensation is politically
difficult
„
Teacher unions hard for the “Terminator” in the US;
Pakistan is no different
„ Use teacher posting and transfers to complement
private sector location decisions
„
„
Access for the poor
Creating the teachers for tomorrow (secondary
schooling is mostly government)
„ Allow districts to set own salaries
„ With recourse to a set of teachers from the province
Experimenting with vouchers
„ Need
Parents are informed
„ Parents care about what governments care
about
„ The supply curve of private schools is
sufficiently elastic
„
„ #1 is probably not a problem, #2 and #3 are
And government schools?
„ Given parental preferences (distance, higher
investments in certain children)
„ Given supply constraints in the private sector
„ Given core failing government schools
„ Need at least an interim strategy to improve
government schools
This is all about teachers, teachers, teachers
„ Hints from US reforms? (NCLB)
„
Oh….and the WB project
„ Private sector share expanded fastest in
Punjab during the 4 years of the WB project
compared to other provinces
„
Despite all funding to government schools only
„ Test-Scores stagnated
„
For both public and private schools
„ The private-public infrastructure gap widened
„
Although a large part of the lending program is
improving infrastructure