No Lost Generation: Supporting the School Participation of Displaced Syrian Children in Lebanon

Figures & data

Map 1. Pilot households and comparison households.

Table 1. Household demographic and composition characteristics at baseline

	Control		Treatment		Balance Test
Variable	Mean	N1	Mean	N2	Diff	Significance
Household size	6.07	497	6.35	590	−0.04	–
# Children (5–14) per household	2.56	497	2.70	590	0.13	–
# Young children (5–9) per household	1.48	497	1.45	590	0.09	–
# Older children (10–14) per household	1.08	497	1.25	590	0.04	–
Time displaced (years)	3.92	495	3.97	590	0.20	–

Notes: Standard errors are robust to heteroscedasticity and clustered at the school level. Differences are estimated based on equation (1), without covariates. All outcomes measure the proportion of households unless otherwise indicated. Significance indicated at three levels: < 0.10, < 0.05, and < 0.01. No significant difference indicated by a dash.

Table 2. Adults ages 17+: demographic characteristics at baseline

	Control		Treatment		Balance Test
Variable	Mean	N1	Mean	N2	Diff	Significance
Female	0.53	1,172	0.52	1,477	0.02	–
Age (years)	35.03	1,172	35.16	1,471	−0.41	–
Literate	0.87	1,170	0.82	1,463	−0.11	< 0.01
Employed (last 7 days)	0.26	1,170	0.25	1,428	−0.05	–
Good health	0.80	1,167	0.61	1,464	−0.01	–
Able to walk	0.61	1,167	0.50	1,464	−0.09	–
Able to sweep	0.63	1,167	0.58	1,461	0.01	–

Notes: Standard errors are robust to heteroscedasticity and clustered at the school level. Differences are estimated based on equation (1), without covariates. All outcomes measure the proportion of adults unless otherwise indicated. Significance indicated at three levels: < 0.10, < 0.05, and < 0.01. No significant difference indicated by a dash.

Table 3. Housing quality and access to basic services at baseline

Variable	Control		Treatment		Balance Test
	Mean	N1	Mean	N2	Diff	Significance
Lives in a house	0.75	497	0.64	590	−0.18	–
Lives in a tent	0.09	497	0.17	590	0.04	–
Dwelling in crowded location*	0.49	497	0.34	590	−0.03	–
Dwelling in poor conditions*	0.11	497	0.12	590	−0.08	–
Dwelling is in dangerous condition*	0.03	497	0.06	590	0.06	< 0.05
Electric energy source	0.83	495	0.92	588	−0.01	–
Access to pumped water	0.40	497	0.33	590	−0.11	–
Access to toilet or latrine	0.99	497	1.00	590	0.01	–

Notes: Standard errors are robust to heteroscedasticity and clustered at the school level. Differences are estimated based on equation (1), without covariates. All outcomes measure the proportion of households unless otherwise indicated. Significance indicated at three levels: < 0.10, < 0.05, and < 0.01. No significant difference indicated by a dash. *Quality was judged by enumerator-assessed perception of the housing.

Table 4. Household economic well-being and exposure to aid at baseline

Variable	Control		Treatment		Balance Test
	Mean	N1	Mean	N2	Diff	Significance
Total income PC USD (past 30 days)	54.3	497	50.1	590	−6.77	–
Not enough food (past 7 days)	0.32	496	0.55	590	0.14	–
Anyone skipped meal (past 7 days)	0.11	496	0.42	590	0.17	< 0.05
Children skipped meal (past 7 days)	0.03	496	0.13	590	−0.02	–
Any assistance (share of households)	0.98	494	0.99	545	0.01	–
MPCA/LCC assistance	0.35	494	0.40	545	0.07	–
WFP assistance	0.91	494	0.92	545	0.00	–
UNICEF winter	0.10	494	0.19	545	0.01	–
UNHCR winter (US$147)	0.57	494	0.59	545	−0.03	–
UNHCR winter (US$75)	0.18	494	0.26	545	0.11	< 0.10

Notes: Standard errors are robust to heteroscedasticity and clustered at the school level. Differences are estimated based on equation (1), without covariates. All outcomes measure the proportion of households unless otherwise indicated. Significance indicated at three levels: < 0.10, < 0.05, and < 0.01. No significant difference indicated by a dash. HH = household. LCC = Lebanon Cash Consortium. MPCA = Multipurpose Cash Assistance Programme.

Table 5. Children ages 5–14: demographic characteristics, education, and time use at baseline

Variable	Control		Treatment		Balance Test
	Mean	N1	Mean	N2	Diff	Significance
Female	0.49	1,237	0.48	1,530	0.01	–
Age (years)	9.10	1,237	9.25	1,530	−0.12	–
Enrolled in school	0.62	1,237	0.65	1,530	0.00	–
Boys	0.63	630	0.64	791	0.00	–
Girls	0.61	607	0.66	739	0.01	–
Young (5–9)	0.65	714	0.64	828	−0.03	–
Old (10–14)	0.59	523	0.66	702	0.05	–
Enrolled in afternoon shift	0.39	1,237	0.30	1,530	−0.02	–
Boys	0.39	630	0.29	791	0.03	–
Girls	0.39	607	0.31	739	−0.06	–
Young (5–9)	0.40	714	0.31	828	−0.06	–
Old (10–14)	0.37	523	0.29	702	0.04	–
Conditional on afternoon shift enrolment
Education expenditure	80.23	483	62.01	457	−25.77	< 0.10
Boys	76.35	245	58.35	227	−14.93	–
Girls	84.24	238	65.62	230	−38.09	< 0.10
Young (5–9)	72.45	289	59.75	256	−6.10	–
Old (10–14)	91.83	194	64.89	201	−53.09	< 0.05
Commutes by bus	0.81	483	0.59	457	−0.06	–
Boys	0.84	245	0.55	227	−0.12	–
Girls	0.79	238	0.63	230	0.01	–
Young (5–9)	0.81	289	0.61	256	−0.08	–
Old (10–14)	0.81	194	0.57	201	−0.03	–

Notes: Standard errors are robust to heteroscedasticity and clustered at the school level. Differences are estimated based on equation (1), without covariates. All outcomes measure the proportion of children unless otherwise indicated. Significance indicated at three levels: < 0.10, < 0.05, and < 0.01. No significant difference indicated by a dash.

Figure 1. Baseline balance.

Notes: The horizontal axis represents distance (in kilometres) to the border separating the pilot governorates (Akkar and Mount Lebanon) from the comparison governorates (North, South, and Nabatieh). The vertical red line represents the border. The vertical axis measures the outcome variable of interest. Dots represent local averages, scaled by the number of households. Linear ordinary least squares regression lines have been fitted to the left and the right of the border. Grey areas represent the 95 per cent confidence interval around the regression line.

*Panels c and d represent data conditional on enrolment in an afternoon shift. Observations, panels a and b: 2,767. Observations, panels c and d: 940.

Table 6. Regression results

				Conditional on second shift school attendance
	Enrolled in school	Enrolled in afternoon shift	In Min/Ila	Education expenditure (USD)	Commutes by bus	Days attended school (past week)
Indicator	(1)	(2)	(3)	(4)	(5)	(6)
Subgroup
All	0.046	−0.003	0.518***	74.4**	0.138**	0.607***
	(0.049)	(0.076)	(0.075)	(37.0)	(0.066)	(0.176)
Boys	0.044	0.025	0.511***	68.1*	0.133*	0.649***
	(0.060)	(0.086)	(0.080)	(39.5)	(0.076)	(0.168)
Girls	0.055	−0.031	0.518***	81.8**	0.144**	0.552***
	(0.048)	(0.078)	(0.079)	(37.2)	(0.065)	(0.201)
Young (5–9)	0.041	−0.006	0.510***	88.5**	0.160**	0.684***
	(0.042)	(0.096)	(0.078)	(38.2)	(0.065)	(0.164)
Old (10–14)	0.063	0.012	0.535***	53.5	0.101	0.511**
	(0.078)	(0.086)	(0.081)	(40.2)	(0.083)	(0.212)
Observations
All	2,767	2,767	2,767	1,392	1,392	1,392
Boys	1,421	1,421	1,421	701	701	701
Girls	1,346	1,346	1,346	691	691	691
Young	1,542	1,542	1,542	855	855	855
Old	1,225	1,225	1,225	537	537	537
End line control group means
All	0.791	0.475	0.021	117.8	0.850	3.806
Boys	0.787	0.471	0.021	119.8	0.848	3.815
Girls	0.794	0.478	0.021	115.8	0.852	3.797
Young	0.898	0.525	0.021	114.1	0.861	3.739
Old	0.644	0.405	0.021	124.4	0.830	3.925

Notes: Standard errors (in parentheses) are robust to heteroscedasticity and clustered at the school level. Differences are estimated based on equation (1), with covariates. All outcomes measure the proportion of children unless otherwise indicated. ***p < .01; **p < .05; *p < .1.

Figure 2. End line school enrolment.

Notes: The horizontal axis represents distance (in kilometres) to the border separating the pilot governorates (Akkar and Mount Lebanon) from the comparison governorates (North, South, and Nabatieh). The vertical red line represents the border. The vertical axis measures the outcome variable of interest. Dots represent local averages, scaled by the number of households. Linear ordinary least squares regression lines have been fitted to the left and the right of the border. Grey areas represent the 95 per cent confidence interval around the regression line. Observations: 2,767.

Figure 3. End line NLG/Min Ila Programme beneficiary households.

Source: Calculated based on administrative data.

Notes: The horizontal axis represents distance (in kilometres) to the border separating the pilot governorates (Akkar and Mount Lebanon) from the comparison governorates (North, South, and Nabatieh). The vertical red line represents the border. The vertical axis measures enrolment in the NLG/Min Ila Programme at follow-up. Dots represent local averages, scaled by the number of households. Linear ordinary least squares regression lines have been fitted to the left and the right of the border. Grey areas represent the 95 per cent confidence interval around the regression line. Observations: 2,767.

Figure 4. End line education outcomes, conditional on enrolment in an afternoon shift.

Notes: The horizontal axis represents distance (in kilometres) to the border separating the pilot governorates (Akkar and Mount Lebanon) from the comparison governorates (North, South, and Nabatieh). The vertical red line represents the border. The vertical axis measures the outcome of interest. Dots represent local averages for clusters of households, scaled by the number of households. Linear ordinary least squares regression lines have been fitted left and right of the border. Grey areas represent the 95 per cent confidence interval around the regression line. In panel a, 28 observations with values above US$400 were dropped for ease of comprehension. Observations, panel a: 1,364. Observations, panels b and c: 1,392.

Figure A1. Attrition of children ages 5–14 from baseline to follow-up.

Notes: The horizontal axis represents distance (in kilometres) to the border separating the pilot governorates (Akkar and Mount Lebanon) from the comparison governorates (North, South, and Nabatieh). The vertical red line represents the border. The vertical axis measures the percentage of households that remain in the sample at follow-up. Dots represent local averages, scaled by the number of households. Linear ordinary least squares regression lines have been fitted to the left and the right of the border. Grey areas represent the 95 per cent confidence interval around the regression line. Observations: 2,897.

Table C1. Robustness tests

Indicator	Enrolled in school	Enrolled in afternoon shift	In Min/Ila	Conditional on second shift school attendance
				Education expenditure	Commutes by bus	Days attended school (past week)
Robustness test
1. Flexible functional form	0.041	−0.012	0.498***	75.3**	0.130*	0.538**
	(0.050)	(0.074)	(0.069)	(37.5)	(0.069)	(0.210)
2. Direct comparison	0.028	0.004	0.526***	72.3**	0.087*	0.506***
	(0.030)	(0.048)	(0.046)	(28.6)	(0.049)	(0.139)
Observations
1	2,767	2,767	2,767	1,392	1,392	1,392
2	2,031	2,031	2,031	1,062	1,062	1,062

Notes: Standard errors (in parentheses) are robust to heteroscedasticity and clustered at the school level. In robustness test 1, the outcome variables are regressed on the control variables used also in the main regressions: child age dummies, child gender, and an indicator taking the value 1 if the child’s parents had studied beyond primary school. Discontinuities in the residuals of these regressions are then estimated using the Stata RD command written by Austin Nichols. The analysis relies on rectangle kernel regressions and a bandwidth restriction of 20 km. In robustness test 2, the observations are first restricted to those living within 20 km of the borders separating pilot and non-pilot governorates. Next, the outcome variables are regressed only on the control variables used in the main regressions and the indicator for living in a pilot governorate. ***p < .01; **p < .05; *p < .1.