El costo de oportunidad de estudiar
•
Las personas sacrifican ingreso al estar en la escuela (forgone earnings):
–
Es más grande que el costo directo
–
Abandonar la escuela es una decisión
–
Dado que estudiar paga, debe ser que no tienen financiamiento
–
Hay espacio para ganancias en eficiencia (y equidad)
•
Para los jóvenes no es fácil pedir prestado:
–
Son menores de edad, sin historial crediticio
–
Contratos difíciles de estructurar y hacer cumplir
–
Fuera de la familia hay grandes asimetrías de información
•
Una opción son las transferencias condicionales de efectivo (CCT):
–
“Te doy $X si haces Y, el cual voy a verificar que hagas”
Tres estudios sobre CCT
Autores
Programa y país
Estructura
Tipo de evaluación
Resultado
Dearden et
al. 2009
Education
Maintenance
Allowance
(EMA), Reino
Unido
Estudiantes de segundo y
tercero de preparatoria, de
familias pobres, 45 o 60
USD por semana
Experimental en un
piloto, con matching
en distintas áreas.
4.5 pp en segundo
año y 6.7 pp en
tercero. Efectos
heterogéneos.
Cameron
2009
Social Safety Net
Scholarships
Program,
Indonesia
Estudiantes de primaria,
secundaria y preparatoria,
familias pobres con alto
riesgo de deserción de
acuerdo con un comité
local, 10, 20 ó 30 USD al
año
Datos de encuestas.
MCO (OLS)
controlando por
selección en variables
observables.
Reducción de la
deserción en
secundaria de 3 pp,
de 10.9 a 7.9%.
Berhman, et
al. 2011
PROGRESA
Oportunidades,
México
Estudiantes de 15 a 21,d e
familias pobras. CCT al
hogar (la madre), junto con
otros beneficios (salud,
nutrición), entre 8 y 55
Aleatorización,
desfasamiento,
DIDM, introducción
desfasada del
programa, usando
.5 a 1 año
adicionales de
escolaridad.
PROBEMS
•
Después de PROGRESA, el gobierno mexicano puso en marcha más CCT
–
Mismo principio, con algunas diferencias de instrumentación
•
¿Cómo evaluarlo?
–
No se hizo aleatorización ex ante como en PROGRESA
–
Retos legales y éticos para hacer un experimento
–
La estructura del programa provee un marco cuasi experimental
•
Regresión discontinua (RDD)
•
Diferencias en diferencias (DD)
•
Los resultados son dramáticos:
–
Impacto muy grande
Introduction
PROBEMS:
•
CCT to high school students in good standing
•
It started in 2007 and currently has over 260k beneficiaries
•
It does not overlap with other programs
•
Means tested:
–
Once enrolled, students fill out a questionnaire
–
Program officials build socioeconomic score
–
Based on availability of funds place cut-off point
•
Different amounts of CCT depending on: gender, GPA, grade, type of school
•
Good standing is verified once or twice a year
•
Discontinuities provide quasi-experimental settings for impact evaluation:
–
Extensive margin: beneficiary v non-beneficiary
Table 1. PROBEMS Cash Transfers, pesos per month
Grade / GPA in previous academic
year/level
Regular High School Technological High School Male Female Male Female
10th grade 6.0 to 7.9 500 525 500 525 8.0 to 10.0 625 650 650 700 11th grade 6.0 to 7.9 525 575 525 575 8.0 to 8.9 650 700 700 750 9.0 to 9.4 850 900 850 900 9.5 to 9.7 900 950 900 950 9.8 to 10.0 950 1000 950 1000 12th grade 6.0 to 7.9 575 625 575 625 8.0 to 8.9 700 750 750 790 9.0 to 9.4 850 900 850 900 9.5 to 9.7 900 950 900 950 9.8 to 10.0 950 1000 950 1000
Source: PROBEMS. The transfers are given 10 months of the academic year. Grades range between 0 and 10, and 6.0 is the minimum passing grade. Technological high schools grant a high school diploma together with a technical degree.
CCT
recipient
status
266,000
Cut-off determined ex post
given funds available and
applicants’ characteristics
Institutional Background
Data
Administrative records from applicants in academic years 2010-11 & 2011-12:
•
~15k beneficiaries right above the cut
•
~15k non-beneficiaries right below the cut
Definition of dropouts:
•
10
thand 11
thgrades: did not enroll the following academic year (September)
•
12
thgrade: did not graduate in the summer
Technical high schools:
•
~2/3 of applicants
•
Administered by the Federal Government
•
SEP can provide records for CCT non-recipients
CCT
recipient
status
266,000
Data
1
30k valid applicants
from technical high
schools around cut-off
.2 .4 .6 .8 1
Fraction of CCT recipients by percentile of socieconomic score
Data
0 .2 .4 .6 .8 1
Dropout rates by CCT recipient status and percentile of socieconomic score
Data
0 .2 .4 .6 .8 1 200 250 300 350 400 450 500 550 600 650 700
Dropout rates by CCT recipient status and household income
Data
0 .2 .4 .6 .8 1
Dropout rates by CCT recipient status and GPA, 10th grade applicants
Data
Empirical Strategy
Sharp or fuzzy RDD?
•
It makes no difference
Three specifications:
•
OLS on full sample (linear model for comparability with IV)
•
OLS ommitting fuzzyness
•
IV on full sample, IV = 1(HH income per capita ≥ $500)
Intensive margin:
•
Difference-in-difference
No evidence of sorting:
Empirical Strategy
Table 2. Regression of covariates on income and CCT recipient status
HH inc. p.c. 1,000s CCT recipient Constant R-squared Piped water 0.119 *** 0.008 0.924 *** 0.003 Water inside the house 0.151 *** -0.011 0.837 *** 0.005 Bath for the family 0.099 *** -0.008 0.815 *** 0.002 Running water 0.213 *** -0.015 0.777 *** 0.007 Electricity 0.017 * 0.001 0.991 *** 0.000 Blender 0.219 *** 0.025 *** 0.845 *** 0.004 Stove 0.106 *** -0.001 0.950 *** 0.005 Radio 0.171 *** 0.018 0.474 *** 0.001 Sewing machine 0.072 ** 0.010 0.101 *** 0.000 DVD player 0.260 *** 0.004 0.324 *** 0.005 Fan 0.098 ** -0.018 0.446 *** 0.002 AC -0.002 -0.032 *** 0.072 *** 0.005 Antenna 0.054 * -0.005 0.084 *** 0.001 Motorcycle 0.014 0.004 0.014 *** 0.000 Truck/Pickup truck 0.029 0.003 0.066 *** 0.000 Microwave oven 0.145 *** -0.026 ** 0.232 *** 0.006 Refrigerator 0.167 *** 0.001 0.908 *** 0.006 Water heater 0.225 *** -0.027 ** 0.292 *** 0.009 Stereo 0.154 *** 0.001 0.176 *** 0.003 TV set 0.125 *** 0.005 0.936 *** 0.004 Washing machine 0.250 *** -0.021 * 0.372 *** 0.008
Results
Table 3. Impact of the CCT on the probability of dropping out 10th grade Explanatory variable Males Females OLS Full sample OLS no fuzzy obs. IV Full sample OLS Full sample OLS no fuzzy obs. IV Full sample CCT recipient -0.491 *** -0.493 *** -0.486 *** -0.562 *** -0.562 *** -0.569 *** Household income per capita 0.000 0.000 0.000 0.000 0.000 0.000 GPA in previous year -0.091 *** -0.089 *** -0.091 *** -0.108 *** -0.108 *** -0.108 *** Dummy for GPA ≥ 8.0 -0.010 -0.013 -0.009 0.004 0.005 0.002 Dummy for GPA ≥ 8.0 x CCT -0.022 -0.020 -0.025 0.032 0.029 0.036 R-square 0.408 0.410 0.408 0.421 0.424 0.421 Obs. 8519 8403 8519 10195 10074 10195
* p-value < 0.05; ** p-value < 0.01; *** p-value < 0.001. Robust standard errors were computed. In the full sample 49.2% of males and 50.9% of females received the CCT.
All specifications include high school fixed effects and controls for household characteristics.
Fuzzy observations are those of students : (1) receiving CCT with a monthly household income per capita of $500 or higher, or (2) not receiving the CCT with a monthly household income per capita lower than $500. They represent 1.35% of male sample and 1.19% of female sample.
Results
Table 3. Impact of the CCT on the probability of dropping out 11th grade Explanatory variable Males Females OLS Full sample OLS no fuzzy obs. IV Full sample OLS Full sample OLS no fuzzy obs. IV Full sample CCT recipient -0.619 *** -0.619 *** -0.636 *** -0.604 *** -0.601 *** -0.592 *** Household income per capita 0.000 0.000 0.000 0.000 0.000 0.000 GPA in previous year -0.082 *** -0.079 *** -0.082 *** -0.083 *** -0.084 *** -0.083 *** Dummy for GPA ≥ 8.0 -0.009 -0.016 -0.013 -0.018 -0.020 -0.014 Dummy for GPA ≥ 9.0 -0.021 -0.013 -0.021 -0.017 -0.009 -0.017 Dummy for GPA ≥ 9.5 0.001 -0.013 0.001 0.047 0.038 0.047 Dummy for GPA ≥ 9.8 -0.057 -0.048 -0.056 -0.048 -0.041 -0.048 Dummy for GPA ≥ 8.0 x CCT 0.022 0.026 0.031 0.009 0.011 0.002 Dummy for GPA ≥ 9.0 x CCT 0.092 * 0.076 0.092 * 0.080 * 0.073 * 0.080 * Dummy for GPA ≥ 9.5 x CCT -0.004 0.009 -0.004 -0.010 -0.001 -0.010 Dummy for GPA ≥ 9.8 x CCT 0.103 0.099 0.104 0.057 0.049 0.057 R-square 0.511 0.515 0.511 0.471 0.473 0.471 Obs. 2871 2832 2871 4093 4032 4093
* p-value < 0.05; ** p-value < 0.01; *** p-value < 0.001. Robust standard errors were computed. In the full sample 50.8% of males and 50.7% of females received the CCT.
Results
Table 3. Impact of the CCT on the probability of dropping out 12th grade Explanatory variable Males Females OLS Full sample OLS no fuzzy obs. IV Full sample OLS Full sample OLS no fuzzy obs. IV Full sample CCT recipient -0.896 *** -0.895 *** -0.897 *** -0.876 *** -0.871 *** -0.868 *** Household income per capita 0.000 0.000 0.000 0.000 0.000 0.000 GPA in previous year -0.036 * -0.036 * -0.036 * -0.037 ** -0.037 ** -0.037 ** Dummy for GPA ≥ 8.0 0.044 * 0.043 * 0.044 * 0.037 * 0.038 * 0.039 ** Dummy for GPA ≥ 9.0 0.040 * 0.040 * 0.040 * 0.023 0.023 0.022 Dummy for GPA ≥ 9.5 0.002 0.001 0.002 0.043 ** 0.044 ** 0.043 ** Dummy for GPA ≥ 9.8 0.036 0.038 0.036 -0.003 -0.003 -0.003 Dummy for GPA ≥ 8.0 x CCT -0.080 *** -0.081 *** -0.079 *** -0.085 *** -0.087 *** -0.090 *** Dummy for GPA ≥ 9.0 x CCT -0.010 -0.010 -0.010 0.017 0.017 0.017 Dummy for GPA ≥ 9.5 x CCT -0.010 -0.009 -0.010 -0.050 * -0.050 * -0.050 * Dummy for GPA ≥ 9.8 x CCT -0.009 -0.010 -0.009 -0.008 -0.009 -0.008 R-square 0.912 0.911 0.912 0.915 0.914 0.915 Obs. 1877 1851 1877 2752 2702 2752
* p-value < 0.05; ** p-value < 0.01; *** p-value < 0.001. Robust standard errors were computed. In the full sample 49.7% of males and 51.1% of females received the CCT.
All specifications include high school fixed effects and controls for household characteristics.