Table 1

Applicant Household Characteristics at Endline

Polynomial RegressionLocal Linear Regression
Order: 1Order: 2Bandwidth = 6Bandwidth = 3Bandwidth = 12
Gender (n = 2973)−0.051*
(0.027)
−0.044
(0.033)
−0.055
(0.039)
−0.032
(0.060)
−0.049
(0.029)
Age (n = 2971)−0.038
(0.073)
0.054
(0.093)
−0.100
(0.117)
0.048
(0.160)
−0.012
(0.085)
Household head education (n = 2706)0.276
(0.225)
−0.087
(0.251)
−0.366
(0.317)
−0.381
(0.466)
−0.003
(0.233)
Household spouse education (n = 2013)−0.235
(0.166)
−0.425*
(0.220)
−0.484*
(0.260)
−0.822**
(0.376)
−0.338
(0.212)
Household highest education (n = 2884)0.137
(0.216)
−0.068
(0.256)
−0.194
(0.325)
−0.175
(0.442)
−0.008
(0.241)
Household mean education (n = 2884)0.096
(0.154)
−0.141
(0.184)
−0.375
(0.230)
−0.400
(0.320)
−0.055
(0.169)
Number of children 6-11 (n = 2973)0.004
(0.047)
−0.027
(0.061)
−0.046
(0.072)
−0.002
(0.123)
−0.026
(0.055)
Number of boys 6-11 (n = 2973)−0.036
(0.031)
−0.037
(0.044)
−0.033
(0.053)
0.020
(0.081)
−0.035
(0.041)
Number of girls 6-11 (n = 2973)0.040
(0.032)
0.010
(0.042)
−0.013
(0.048)
−0.021
(0.078)
0.009
(0.038)
Number of children 12-18 (n = 2973)−0.016
(0.053)
0.037
(0.070)
0.078
(0.081)
0.069
(0.129)
0.032
(0.062)
Number of boys 12-18 (n = 2973)−0.064
(0.054)
−0.075
(0.061)
−0.080
(0.073)
−0.077
(0.097)
−0.065
(0.056)
Number of girls 12-18 (n = 2973)0.048
(0.046)
0.111*
(0.058)
0.158**
(0.065)
0.147
(0.111)
0.097*
(0.054)
Number of adults 18-25 (n = 2973)−0.012
(0.054)
0.033
(0.072)
0.044
(0.087)
−0.048
(0.146)
0.028
(0.067)
Number of adults >25 (n = 2973)−0.046
(0.051)
−0.077
(0.067)
−0.081
(0.088)
−0.173
(0.132)
−0.057
(0.062)
  • Notes: Coefficients and standard errors from regression of a given characteristic on indicator variable for children with dropout-risk score above the cutoff. Polynomial regression order 1 includes dropout-risk score and its interaction with the indicator for children with scores above the cutoff; polynomial regression order 2 includes dropout-risk score and its square, and interactions with the indicator for children with scores above the cutoff; choice of bandwidth for Local Linear Regressions described in body of paper. Standard errors cluster at level of primary feeder school. Highest / mean education calculated by taking the maximum / average of education in years per household, only including household members over age 25.

  • *** p-value < 0.01,

  • ** p-value < 0.05,

  • * p-value < 0.1.