we arjun bedi aiid_jp_asb
TRANSCRIPT
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 1/24
The impact of a cash transfer program on cognitiveachievement: The Bono de Desarrollo Humano
of Ecuador
Juan PonceFlacso, Ecuador
Arjun Singh BediInternational Institute of Social Studies, Erasmus University Rotterdam
Ocotber 2010
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 2/24
• Throughout Latin America, conditional cash transfer (CCT)programs play an important role in social policy
• CCT programs started in the 1990s and aim to influence humancapital accumulation and serve as a means of breaking the inter-generational cycle of poverty
• Provide a cash transfer to a target group conditional on certain actions
Visits to health centres – growth monitoring and nutritional supplements
Enrol in school and attend school regularly
Motivation and Introduction
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 3/24
• A number of CCT programs operate in Latin America and they have been systematically evaluated
• On the education front, a majority of studies find that CCTprograms boost school enrolment and ensure regular schoolattendance
• These are clearly the first steps to ensure higher educationalattainment and achievement
• However, a focus on enrolment may not be enough to ensureaccumulation of adequate human capital
• It is important to examine the effect of such programs oncognitive achievement
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 4/24
• Higher cognitive achievement as captured by testscores, are likely to ensure a longer duration in school
and are also correlated with labour market success US data – Murnane, Willet and Levy (1995); Jencks
and Phillips (1999); Rose (2006)
Developing countries – Alderman et al. (1996);Lavy et al. (1997)
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 5/24
• Theoretically, such programs are likely to influence cognitiveachievement in several ways
• Positive effects through•
Increase/ensure regular attendance and this in turn is likely to
translate into higher test scores•
Increase household incomes, increased food consumption, betternutrition
•
Reductions in child labour both in terms of the probability of
working and hours of work
•
There may also be negative effects
• Increases in school enrolment may translate into congestedclassrooms
•
If the program encourages less able students to enrol, then changesin student composition may lead to a decline in average test scores
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 6/24
• Studies evaluating the effect of CCT programs on cognitiveachievement are relatively scarce
• Contribution of the paper is to evaluate the effect of theEcuadorian cash transfer program on students’ cognitiveachievement
• We exploit the manner in which the program/Bono is allocatedand rely on a regression discontinuity design to identify theimpact of the program on second grade cognitive achievement
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 7/24
CCT programs in Latin America
• Started in Brazil in 1995, followed by Mexico in 1997,Honduras in 1998 and Nicaragua in 2000
• Programs have been heavily evaluated along several
dimensions School enrolment (Behrman, et al. 2005; Schultz, 2004;
Skoufias, 2000)
Attendance (Maluccio and Flores, 2004)
Nutrition, illness (Skoufias, 2000; Maluccio and Flores, 2004)
Child work (Duryea and Morrison, 2004)
Cognitive outcomes (Behrman, Sengupta and Todd, 2000)for Mexico, find no effect on test scores
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 8/24
Ecuador’s program
•
BDH program was launched in 2003 and incorporated
two existing programs• Main objective is to improve the formation of human
capital amongst poor families in Ecuador
• Education and health component Education component requires children from age 6-15 to
enrol in school and attend at least 90 percent of school days
Health component focuses on children under 6• Cash transfer of USD 15 per month per family
•
Targeted at families with average monthly expenditureof USD 100
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 9/24
Ecuador’s program• Selection of beneficiaries
Program uses an individual targeting strategy
Participation is based on an index called Selben (“Sel”ection of “Ben” ificiaries)
Index is constructed on the basis of 27 variables (household
infrastructure, educational characteristics, assets) and takes values from 0 to 100
Those scoring below 50.65 are eligible to receive benefits
In 2004, the annual budget was USD 190 million, around 1percent of GDP and covered 1.1 million households or 40percent of the population
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 10/24
Ecuador’s program• Vos et al. (2001) find a 5 percentage point increase in school
enrolment
• Leon and Younger (2007) find small nutritional effects
• Schady and Araujo (2006) report increases in school enrolmentin the age group 6-17 (about 10 percentage points) and reduction
in child work (about 17 percentage points)• Ponce (2008) refines this finding
Effects are heterogeneous
Restricted to children from the bottom quintile while for families close tothe cut off there are no enrolment effects
Program beneficiaries around the neighbourhood of the eligibility cut off experience sharp increases in food (25 percent) and educational
expenditure (46 percent)
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 11/24
Empirical strategy • To isolate the effect of the BDH on students’ test score we begin
with the following educational production function
• Y i
- outcome variable; S i
is the Selben index ; T i
indicates receipt
of BDH
• Program participation is not random and it is quite likely that thecoefficient on T i
is likely to be downward biased
ii
j
i
j
jii uT S X Y
3
1
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 12/24
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 13/24
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 14/24
• If individuals were assigned to treatment solely on thebasis of the assignment variable then it may be arguedthat the program allocation rule replicates randomassignment around the cut off point
• In this case it is unlikely that program participation isdetermined so strictly
• There is a fair degree of “fuzziness”
• Treatment status depends on the Selben index but in astochastic manner
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 15/24
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 16/24
• Since we have a variable (an instrument, Z i ) that is, the cut off point which determines treatment status
• There is no reason to expect that this cut off point is correlated with u i
• Estimates based on (3) provide a local average treatment effect
)1(3
1
ii
j
i
j
jii uT S X Y
)2(
3
1
ii
j
ji jii w Z S X T
)3(
3
1uT S X Y i
j
j
i jii
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 17/24
• Data were collected between November 2004 and February 2005
• A year and a half after the program had been launched• The data cover rural areas of the country and the capital• Three different instruments
Standardized tests in mathematics and language were conducted forstudents in 2nd
and 4th
grades
School and teacher characteristics
Household information
• Second grade sample has 2588 children
1469 in treatment and 1119 in the control group
Data
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 18/24
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 19/24
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 20/24
Validity of the RDD
•
First stage estimates (equation 2)
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 21/24
2
1 3
1 0
7
1 9
2 8 3
0
5 2
3 7
5 3
5 8
6 4
5 4
5 1
6 2
5 8
5 2
5 5
5
0
5 5
5 3 5
4
5 8
6 7
5 3
5 3
5 2
5 2
5 5
6 4
7 5
7 4
5 9
6 2
5 8
5 8
7
2
5 3
5 3
8 0
6 9
5 8
4
9
4 2
4 2
4 7
3 3 3
5
3 0
2 6
3 1
1 7
2 5
1 5 1
6
2
9
1 2
5
8
4
1 2
1 1
0
2 0
4 0
6 0
8 0
F r e q u e n c y
1 0
1 5
2 0
2 5
3 0
3 5
4 0
4 7
. 6 5
5 0
. 6 5
5 3
. 6 5
6 0
6 5
7 0
7 5
Selben index
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 22/24
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 23/24
8/6/2019 We Arjun Bedi Aiid_jp_asb
http://slidepdf.com/reader/full/we-arjun-bedi-aiidjpasb 24/24
• CCT programs appear to have a large effect on enrolment,attendance, reduction of child labour
• These are the first steps but a focus on learning may also berequired
• We exploited the program’s design and used an arguably credible empirical strategy and found that the BDH does notlead to higher rest scores
Maybe we are analyzing the effects too early
Maybe we need to examine those lower down in the Selben
distribution
•
Highlights strengths and limitation of such programs
Concluding remarks