Who benefits from the South African Child Support Grant?: The role of gender and birthweight

Figures & data

Table 1. Summary statistics.

	Treated		Comparison
Variable	Mean	SD	Mean	SD	Mean difference	t-statistic
Delay (in years)	2.05	2.07	7.67	5.39	5.623***	−43.9
Motivation	0.57	0.79	−0.87	0.56	−1.439***	(−59.27)
Employed	0.24	0.43	0.24	0.43	−0.00449	(−0.31)
Married	0.35	0.48	0.41	0.49	0.0541**	−3.28
Primary caregivers’ education	7.36	4.38	6.1	4.79	−1.269***	(−8.17)
Primary caregivers age	38.16	13.65	43.87	17.68	5.706***	−10.87
HH head gender	0.36	0.48	0.35	0.48	−0.00914	(−0.56)
Child’s gender	0.5	0.5	0.5	0.5	−0.00038	(−0.02)
Electricity	0.64	0.48	0.63	0.48	−0.00616	(−0.37)
Water	0.17	0.37	0.19	0.39	0.0245	−1.89
Telephone	0.03	0.16	0.04	0.19	0.00902	−1.52
Toilet	0.2	0.4	0.24	0.43	0.0482***	−3.44
Food expenditure (AES)	183.92	120.82	195.55	151.87	11.63*	−2.54
Child’s age	6.64	3.74	7.86	5.39	1.215***	−7.97
Rural	0.74	0.44	0.72	0.45	−0.0278	(−1.84)
Child’s relationship to grant recipient	0.74	0.44	0.51	0.5	−0.236***	(−14.91)
Informal dwelling	0.08	0.27	0.08	0.26	−0.00334	(−0.37)
Hospital visits (No=1)	0.45	0.5	0.49	0.5	−0.0109	(−1.91)
Observations	2170		1418		3588

t-statistics in parentheses.

*p < 0.05, **p < 0.01, *** p < 0.00.

Table 2. Delay equation.

VARIABLES	Years of delay
Child age	0.787***
	(0.017)
Rural location	−0.309*
	(0.163)
Care giver relationship	−2.053***
	(0.156)
Constant	1.187***
	(0.222)
Observations	3,588

Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 3. First three moments after entropy balancing.

	Treat			Control
	Mean	Variance	Skewness	Mean	Variance	Skewness
Motivation	0.569	0.628	0.660	0.568	0.629	0.795
Employed dummy	0.243	0.184	1.199	0.243	0.184	1.199
Married dummy	0.352	0.228	0.619	0.352	0.228	0.619
*Primary caregivers’ education (<5years)	0.313	0.215	0.805	0.313	0.215	0.805
Primary caregivers’ education (between 5 and 10 years)	0.396	0.239	0.426	0.396	0.239	0.426
*Primary caregivers age (<40years)	0.636	0.232	−0.567	0.637	0.232	−0.568
Primary caregivers age (between 40 and 60 years)	0.277	0.200	0.999	0.276	0.200	1.000
Household head’s gender	0.362	0.231	0.575	0.362	0.231	0.575
Child’s gender	0.505	0.250	−0.018	0.505	0.250	−0.019
Electricity	0.637	0.231	−0.571	0.637	0.231	−0.571
Water	0.166	0.138	1.796	0.166	0.139	1.796
Telephone	0.028	0.027	5.762	0.028	0.027	5.759
Toilet	0.196	0.158	1.533	0.196	0.158	1.532
Food expenditure (AES)	183.9	14598.	3.710	183.9	14601	3.496
Informal dwelling dummy	0.079	0.073	3.127	0.079	0.073	3.126
Hospital visits	0.454	0.248	0.187	0.454	0.248	0.187

Note that by definition any difference between the moments are not significant

*Primary caregivers age and education are feed to the algorithm as dummy variables to ease convergence

Table 4. Effect of CSG with and without low birthweight.

Dependent variable height-for-age score	(1)	(2)
Treatment	0.14	0.15
	(0.18)	(0.18)
Low birthweight		−0.44**
		(0.21)
Constant	−0.21	−0.18
	(0.43)	(0.43)
Other covariates	(included)	(included)
Observations	2,837	2,837
R-squared	0.08	0.08

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

By other covariates I mean the covariates listed in Table 2.

Table 5. Results disaggregated by gender.

	Boys		Girls
Dependent variable height-for-age score	(1)	(2)	(3)	(4)
CSG	−0.03	0.03	0.27***	0.29***
	(0.27)	(0.25)	(0.11)	(0.11)
		−0.44**		−0.38
Low birthweight		(0.21)		(0.35)
Constant	−0.67	−0.84	−1.43***	−1.42***
	(0.57)	(0.56)	(0.18)	(0.31)
Other covariates	(included)	(included)	(included)	(included)
Observations	1,436	1,436	1,401	1,401

Figure 1. Unconditional quantile effect of CSG.

Figure 2. Unconditional quantile regression for boys.

Figure 3. Unconditional quantile regression for girls.

Table 6. Effect of low birthweight.

Dependent variable height-for-age score	(1)
Low birthweight	−0.49***
	(0.19)
Constant	1.50***
	(0.37)
Other covariates	(included)
Observations	2,837
R-squared	0.08

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

Figure 4. Quantile effect of low birthweight.

Table A1. Summary statistics when low birthweight is the treatment variable.

Entropy weighting
	Treat			Control
	Mean	Variance	Skewness	Mean	Variance	Skewness
Treatment	0.70	0.21	−0.85	0.69	0.21	−0.84
Motivation	0.11	0.65	0.76	0.11	0.65	0.76
Employed	0.24	0.18	1.25	0.24	0.18	1.23
Married	0.40	0.24	0.40	0.40	0.24	0.39
Primary caregivers’ education (<5years)	0.27	0.20	1.01	0.28	0.20	1.00
Primary caregivers’ education (between 5 and 10 years)	0.39	0.24	0.44	0.39	0.24	0.43
Primary caregivers age (<40years)	0.70	0.21	−0.85	0.69	0.21	−0.84
Primary caregivers age (between 40 and 60 years)	0.21	0.17	1.46	0.21	0.16	1.44
HH head gender	0.43	0.25	0.28	0.43	0.25	0.27
Child’s gender	0.43	0.25	0.28	0.43	0.25	0.27
Electricity	0.70	0.21	−0.85	0.69	0.21	−0.84
Water	0.27	0.20	1.01	0.28	0.20	1.00
Telephone	0.04	0.04	4.75	0.04	0.04	4.73
Toilet	0.21	0.17	1.46	0.21	0.16	1.44
Food expenditure (AES)	187	15990	1.81	187	15967	1.82
Informal dwelling	0.11	0.10	2.53	0.11	0.10	2.51
Hospital visits	0.35	0.23	0.62	0.36	0.23	0.60

Figure A1. Quantile effect (Conditional quantile regression).

Table A2. Effect of CSG with and without low birthweight when the estimation of motivation excludes the caregiver relationship.

Dependent variable height-for-age score	(1)	(2)
Treatment	0.22	0.21
	(0.18)	(0.18)
Low birthweight		−0.42**
		(0.21)
Constant	−1.20***	−1.23***
	(0.37)	(0.37)
Other covariates	(included)	(included)
Observations	2,837	2,837
R-squared	0.08	0.08

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

By other covariates I mean the covariates listed in Table 2.

Figure A2. Unconditional quantile effect of CSG when the estimation of motivation excludes the caregiver relationship.

Table A3. Results disaggregated by gender when the estimation of motivation excludes the caregiver relationship.

	Boys		Girls
Dependent variable height-for-age score	(1)	(2)	(3)	(4)
CSG	−0.08	−0.14	0.68***	0.72***
	(0.24)	(0.23)	(0.19)	(0.20)
		−0.73***		−0.41
Low birthweight		(0.25)		(0.36)
Constant	−0.42	−0.31	−1.36***	−1.45***
	(0.58)	(0.53)	(0.27)	(0.25)
Other covariates	(included)	(included)	(included)	(included)
Observations	1,436	1,436	1,401	1,401

Figure A3. Unconditional quantile regression for boys when the estimation of motivation excludes the caregiver relationship.

Figure A4. Unconditional quantile regression for girls when the estimation of motivation excludes the caregiver relationship.

Table A4. Effect of low birthweight when the estimation of motivation excludes the caregiver relationship.

Dependent variable height-for-age score	(1)
Low birthweight	−0.49**
	(0.20)
Constant	−1.57***
	(0.38)
Other covariates	(included)
Observations	2,837
R-squared	0.08

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

Figure A5. Quantile effect of low birthweight when the estimation of motivation excludes the caregiver relationship

Table A5. Effect of CSG with and without low birthweight (birthweight<2.3 kg).

Dependent variable height-for-age score	(1)	(2)
Treatment	0.14	0.22
	(0.18)	(0.19)
Low birthweight		−0.58***
		(0.22)
Constant	−0.21	−0.39
	(0.43)	(0.36)
Other covariates	(included)	(included)
Observations	2,837	2,837
R-squared	0.08	0.11

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

By other covariates I mean the covariates listed in Table 2.

Figure A6. Unconditional quantile effect of CSG (birthweight<2.3 kg).

Table A6. Results disaggregated by gender (birthweight<2.3 kg).

	Boys		Girls
Dependent variable height-for-age score	(1)	(2)	(3)	(4)
CSG	−0.04	0.03	0.27**	0.29***
	(0.27)	(0.25)	(0.11)	(0.11)
		−0.44*		−0.61
Low birthweight		(0.24)		(0.46)
Constant	−0.65	−0.84	−1.43***	−1.43***
	(0.56)	(0.55)	(0.22)	(0.23)
Other covariates	(included)	(included)	(included)	(included)
Observations	1,436	1,436	1,401	1,401

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

Figure A7. Unconditional quantile regression for boys (birthweight<2.3 kg).

Figure A8. Unconditional quantile regression for girls (birthweight<2.3 kg).

Table A7. Effect of low birthweight (birthweight<2.3 kg)

Dependent variable height-for-age score	(1)
Low birthweight	−0.41*
	(0.24)
Constant	−1.34***
	(0.46)
Other covariates	(included)
Observations	2,837
R-squared	0.08

Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

Figure A9. Quantile effect of low birthweight when the estimation of motivation excludes the caregiver relationship.