Do the Extra Hours of Tuition Pay off?

advertisement
Presented at IGC Bangladesh Conference
December 20, 2011
DO EXTRA HOURS OF TUITION
PAY OFF?
Atonu Rabbani (Department of Economics, University of Dhaka)
Ummul Hasanath Ruthbah (Department of Economics, University
of Dhaka)
with
Salim Hossain (Department of Psychology, University of Dhaka)
Golam Sarwar (ERG)
PRIMARY SCHOOL EDUCATION IN
BANGLADESH




In Bangladesh 9% of all children, aged between 6
– 10 years, are left out of school.
47% of those who enrol drop out before
completing grade 5.
Very low level of achievements in basic
competency tests.
There is also a huge disparity in the quality of
education.
2
GOVERNMENT POLICY

GoB Policy on primary education are based on World
Education Forum 2000 Framework and MDG 2 and are
reflected in the Bangladesh Education Policy 2010 as:








Providing universal primary education,
Increasing the completion rate of primary education,
Reducing the rate of drop outs,
Reducing the rate of repetition,
Improving the quality of primary education,
Resolving the problem of scarcity of teachers and teaching
materials,
Implementation of unified primary education program,
Extending primary education to grade VIII.
3
SUPPLEMENTARY TUTORING BY CDIP: THE
EDUCATION SUPPORT PROGRAM (ESP)




CDIP has been operating learning centres adjacent to
primary schools in different districts of Bangladesh
since 2005.
These centres provide supplementary tuition (about
10 hours per week) to primary school students in
nursery, grade 1 and grade 2.
Currently it operates through 1,750 learning centres
adjacent to the primary schools.
The goals are:
To improve class performance (i.e. test scores) of grade 1
and 2 students from poor and illiterate households;
 To strengthen the educational foundation of the students
belonging to poor and illiterate households at the entry
level; and
 To reduce the primary school dropout rate in its geographic
areas of operation.

4
SUPPLEMENTARY TUTORING BY CDIP: THE
EDUCATION SUPPORT PROGRAM (ESP)
CDIP has planned to extend this program further
in other parts of the country.
 For further expansion of the ESP, it is important
for CDIP to explore whether this supplementary
tutoring has been achieving its intended goals.
 The objectives / scope of the evaluation study set
by ERG are:

To estimate the short term and long term effect of the
ESP on students’ class performance;
 To estimate the effect of the program on primary
school dropout rates; and

5
METHODOLOGY





The DID approach- estimate the “treatment effect on
the treated”.
Students who were in grade 2 in 2008 and received
the tutoring in CDIP learning centres constitute the
treatment group.
Those who were in grade 2 in 2008 but didn’t receive
tutoring at CDIP learning centres constitute the
control group.
For both the treatment and the control groups the
performance of the students in grade one can be
viewed as the pre treatment observations.
Then their performance in the final exams of grade
two, grade three and grade four can be viewed as the
post treatment observations.
6
METHODOLOGY
Test Scores in Final
Students who
participated in
ESP
(Treatment)
Students who
did not
participate in
ESP (Control)
Difference
between
treatment and
control groups
2007 (grade 1)
2008 (grade 2)
XT2007
XT2008
XC2007
XC2008
XT2007 - XC2007
XT
2008
-
XC2008
Difference in
test scores
between 2007
and 2008
XT2008 - XT2007
XC2008 - XC2007
(XT2008 - XT2007) (XC2008 - XC2007)
= DID estimate
7
SAMPLING STRATEGY
CDIP had 304 learning centres in 2008 operating
in 33 unions of 8 upazilas in Bangladesh.
 Only 262 centres had students from grade 2.
 We would require a total of 1900 observations
(950 in each of treatment and control groups)
selected from 159 learning centres and the
associated primary schools.
 Multistage sampling

Select the learning centres
 Select students who were in grade 2 in 2008 and
participated in the program
 Select control students (6 on average) from the
schools who were in grade 2 in 2008 but did not
participate in the program.

8
THE SURVEYS

There will be three sets of information:
Performance of the treatment and control students in
the final exams in grade one (2007), grade two (2008),
grade three (2009), grade four (2010) and in the first
term in grade five (2011) - to be collected from the
primary schools.
 Background of students in both treatment and
control groups – socio-economic conditions, to be
collected through a household survey.
 School information – to be collected from the primary
schools.

9
THE FIELD EXPERIENCE
It was not possible to follow this sampling
strategy in the field.
 Could not get the complete list of students who
were in grade two in 2008 and attended the ESP.
 We collected data on 2147 students, of whom
1078 students attended 144 different CDIP
learning centers in 2008.
 The schools could provide the marks for 2007
when the students were in class one for only 1215
students.
 Therefore we were forced to use this subsample
for our analysis.

10
DATA: HOUSEHOLD CHARACTERISTICS
Treatment
Control
Age of
Age of HH
Househol
Head’s
d Head
Spouse
(2008)
(2008)
38.25
31.68
8.54
7.70
Std Dev
8.1
7.03
6.54
6.36
N
596
571
592
573
604
Mean
37.99
31.56
8.61
7.98
97.68
Std Dev
7.28
6.16
6.37
6.03
N
604
586
566
538
Mean
HH Heads
Spouse’s
Years of
Years of
Education Education
Male HH
Head%
95.48
597
11
DATA: STUDENT CHARACTERISTICS
All
Treatment
Control
Dropout
Age
% girl
Mean
10.72
56.21
4.69
Std Dev
0.79
N
1198
1215
1215
Mean
10.77
56.62
2.81
Std Dev
0.82
N
596
604
604
Mean
10.67
55.81
6.55
Std Dev
0.75
N
602
611
611
rate%
12
DATA: SCORES OBTAINED IN FINAL EXAMS
Treatment
Grade 1
Grade 2
Grade 3
Grade 4
Grade 5
Mean
203.33
175.76
151.37
146.21
132.9
Std Dev
57.284
54.18
55.21
53.14
48.29
604
584
543
485
504
Mean
205.97
173.83
152.06
148.1
139.52
Std Dev
55.77
53.59
53.99
52.63
50.85
611
584
550
483
487
N
Control
N
13
RESULTS: CLASS PERFORMANCE (DID
ESTIMATES)
Total
Bengali
English
Math
Grade 2
0.72
0.114*
-0.009
0.062
Grade 3
(.06)
.013
(.059)
(.058)
.0001
(.046)
(0.059)
.040
(.063)
(0.068)
-.002
(.063)
Grade 4
.042
.049
.086
-.047
(.065)
(.061)
(.061)
(.072)
-.026
-.011
-.017
-.046
(.057)
(.056)
(.058)
(.059)
Grade 5
14
RESULTS: DROPOUT RATE (LOGIT AND
PROBIT ESTIMATES)
Dependent variable - Treat
Logit estimates
Probit estimates
Coefficient
Average marginal
effect
-.98**
-.04**
(.36)
(.020)
-.44**
-.04**
(.16)
(.019)
15
FIGURE 1: TOTAL MARKS OBTAINED
250
200
Treatment
205.97
Marks (out of 300)
203.33
Control
173.83
152.06
175.76
150
151.37
148.12
139.52
146.21
132.93
100
50
16
0
2007
2008
2009
2010
2011
Year
FIGURE 2: DIFFERENCE IN MARKS
BETWEEN PRE-POST TREATMENT YEARS
5
4.57
4
Marks (out of 300)
3
1.95
2
0.73
1
0
2008
2009
2010
2011
-1
-2
-3
-4
-5
Year
-3.95
17
CONCLUSION



The intervention lowered the drop-rates significantly.
It is possible that such interventions can have a
higher average treatment effect in the population and
scaling up of the program can further give
opportunity to understand this.
The education support program did not exhibit any
significant (statistically or point-wise) impact on test
scores.
The control group chose to receive similar treatments from
other sources (e.g. private tutors).
 There were lots of “good” and privileged students who
definitely biased the impact downward.
 It is difficult to make a proper evaluation ex post. Even
after carefully selecting a sample our study was seriously
constrained by availability of data

18
IMPLICATIONS
After-hour tuitions offered to students did
manage to retain students through grade 5 once
they received the interventions during grade 2.
 Because of selection of schools, the benchmark
drop-out rates were lower than national average.
Yet the intervention lowered the drop-rates
significantly.
 However, it is not through increasing-the-returnsfrom-better-education-by-increasing-test-scores
channel!
 What next?

19
Thank you.
20
Download